Object Invocation Protocol · protocol specification

OIP system: MISC

Posted Jul 2, 2026 · Updated Sep 25, 2026 · 55 min read

MISC

A generated article for one OIP shelf. It lists every operation in this API/CLI/MCP/device/model/core subsystem, links each leaf article, and gives the ledger path for proof.

This page is the operating article for one build subsystem. It is generated from live directory rows. If a task belongs to this subsystem, scan the operations below, open the matching capability article, run only the exact object named there, and verify by receipt.

Kind: core. Capabilities: 225. Machine system map: /api/dispatch?map=MISC&format=markdown. Root: /a/oip.

Operations

SHA256_LOWER

SHA-256 hex of lowercased trimmed input Use when: hash em/ph/external_id for Meta CAPI or any PII hashing Arguments: $1. Human article: /a/oip-capability-sha256-lower. Machine doc: ?key=SHA256_LOWER&format=markdown. Invocation history: /api/invocations?object_id=SHA256_LOWER.

UPPER

Uppercase a string. Use only for cosmetic transforms; not for user-visible reply text Use when: you need to upper Arguments: $1. Human article: /a/oip-capability-upper. Machine doc: ?key=UPPER&format=markdown. Invocation history: /api/invocations?object_id=UPPER.

LOWER

Lowercase a string. Use before SHA256 hashing of emails or other case-insensitive identifiers Use when: you need to lower Arguments: $1. Human article: /a/oip-capability-lower. Machine doc: ?key=LOWER&format=markdown. Invocation history: /api/invocations?object_id=LOWER.

DEDUP_INSERT

Insert message_id into blooio_dedup. Use ONLY from the inbound webhook handler before replying Use when: you need to dedup insert Arguments: $1. Human article: /a/oip-capability-dedup-insert. Machine doc: ?key=DEDUP_INSERT&format=markdown. Invocation history: /api/invocations?object_id=DEDUP_INSERT.

LOG_TAIL

Last 50 kernel log rows. Use to inspect recent dispatch activity (trace + step tree). Use when: When you need to last 50 kernel log rows. Human article: /a/oip-capability-log-tail. Machine doc: ?key=LOG_TAIL&format=markdown. Invocation history: /api/invocations?object_id=LOG_TAIL.

LOG_ASSET

SERVE_PAGE

Body HTML for $1=slug (used by functions/slug].js to serve /<slug>). Use ONLY from the page renderer. Use when: When you need to body HTML for. Human article: [/a/oip-capability-serve-page. Machine doc: ?key=SERVE_PAGE&format=markdown. Invocation history: /api/invocations?object_id=SERVE_PAGE.

ADD_ROW

Insert a new row into the directory (tools, agents, flows). Use when: creating a new tool, agent, or flow from the model. Arguments: key|type|target|auth|content. Everything after the 4th pipe is content.. Human article: /a/oip-capability-add-row. Machine doc: ?key=ADD_ROW&format=markdown. Invocation history: /api/invocations?object_id=ADD_ROW.

EDIT_ROW

Edit an existing directory row (upsert). Use when: changing a tool definition, updating a prompt, or fixing a row. Arguments: key|type|target|auth|content. Everything after the 4th pipe is content.. Human article: /a/oip-capability-edit-row. Machine doc: ?key=EDIT_ROW&format=markdown. Invocation history: /api/invocations?object_id=EDIT_ROW.

EDIT_MEMORY

Change one line of your memory (or any row). $1=key, $2=the exact existing line, $3=the new line Use when: you need to edit memory Arguments: $1 | $2 | $3. Human article: /a/oip-capability-edit-memory. Machine doc: ?key=EDIT_MEMORY&format=markdown. Invocation history: /api/invocations?object_id=EDIT_MEMORY.

DEL_ROW

Delete a directory row. Use when: removing a tool that is broken, stale, or unwanted. Arguments: $1 = the row key.. Human article: /a/oip-capability-del-row. Machine doc: ?key=DEL_ROW&format=markdown. Invocation history: /api/invocations?object_id=DEL_ROW.

VERIFY_BLOOIO_SIG

HMAC-SHA256 hex of $1=body using env.BLOOIO_WEBHOOK_SECRET Use when: verify Blooio inbound signatures Arguments: $1. Human article: /a/oip-capability-verify-blooio-sig. Machine doc: ?key=VERIFY_BLOOIO_SIG&format=markdown. Invocation history: /api/invocations?object_id=VERIFY_BLOOIO_SIG.

VERIFY_CLAIM

Recompute a claim from its committed receipts with its methodology and say match or mismatch. Verification is a capability, not an opinion: anyone may run it, and a better method may enter as another verifier row. Use when: checking a provider's claim before trusting or paying for it. Arguments: claim_id. Human article: /a/oip-capability-verify-claim. Machine doc: ?key=VERIFY_CLAIM&format=markdown. Invocation history: /api/invocations?object_id=VERIFY_CLAIM.

TOOLS_IN

Return JSON {key,type,docs}] of planner-visible tools in $1. $1=category, $2=limit (default 30, max 100). Use as stage 2 of tool selection after CATEGORIES Use when: you need to tools in Arguments: $1 | $2. Human article: [/a/oip-capability-tools-in. Machine doc: ?key=TOOLS_IN&format=markdown. Invocation history: /api/invocations?object_id=TOOLS_IN.

APPS_SCRIPT_RUN

Generic Apps Script invocation. $1=action name, $2=JSON args. Routes to the airunner web app deployment which has Google Drive / Sheets / Tasks / Calendar / Apps Script-runtime access. Use when you need to read/write the user's Google stuff. $$2+ (double dollar, N+) inlines every remaining arg rejoined with | and raw — required because the JSON args blob routinely contains real data (a business name, a drafted email body) with a literal | character in it; plain $2 silently truncates the JSON at that first embedded pipe, which is the true cause of every past 'airunner_body_truncated_in_transit' Use when: any natural language about Google Drive / Sheets / Tasks / Calendar / Apps Script. Arguments: see content. Human article: /a/oip-capability-apps-script-run. Machine doc: ?key=APPS_SCRIPT_RUN&format=markdown. Invocation history: /api/invocations?object_id=APPS_SCRIPT_RUN.

PANEL

Ask a panel of LLMs the same questions about one article. Body JSON: {article, questions], models[]}. Default models GROK_CHAT, KIMI_CHAT, WORKERS_AI_CHAT Use when: you need to panel Arguments: see content. Human article: [/a/oip-capability-panel. Machine doc: ?key=PANEL&format=markdown. Invocation history: /api/invocations?object_id=PANEL.

GEN_DUAL

Generate with BOTH OpenAI gpt-image-1.5 and Grok Imagine; store both to R2; return both links. If reference_url is given, both EDIT it Use when: you need to gen dual Arguments: prompt|reference_url. Human article: /a/oip-capability-gen-dual. Machine doc: ?key=GEN_DUAL&format=markdown. Invocation history: /api/invocations?object_id=GEN_DUAL.

DOCS_GET

Read one writer/build doc from the D1 docs table by slug (style_topology, slot_specs, judge_prompt, arcads, blooio, cloudflare, ai-gateway, 2chat, build-intent). Returns {slug,title,body} Use when: you need to docs get Arguments: $1. Human article: /a/oip-capability-docs-get. Machine doc: ?key=DOCS_GET&format=markdown. Invocation history: /api/invocations?object_id=DOCS_GET.

DOCS_SEARCH

Search stored docs. Arg: query. Returns slugs + snippets Use when: you need to docs search Arguments: $1. Human article: /a/oip-capability-docs-search. Machine doc: ?key=DOCS_SEARCH&format=markdown. Invocation history: /api/invocations?object_id=DOCS_SEARCH.

REPO_SNAPSHOT

Read the FULL repo snapshot from KV: {sha, ts, byte_count, content} — content is ~310KB and floods the next turn's input. Use SNAPSHOT_META for metadata-only. Call this only when you need the raw blob itself Use when: you need to repo snapshot Arguments: none. Human article: /a/oip-capability-repo-snapshot. Machine doc: ?key=REPO_SNAPSHOT&format=markdown. Invocation history: /api/invocations?object_id=REPO_SNAPSHOT.

REPO_ABSORB

Absorb a GitHub repo into directory rows on the owner's voice command. Args: a GitHub URL or owner/repo plus any extra instruction. Hands the job to BUILDER, which clones to ~/_absorbed/, reads README/AGENTS.md, ADD_ROWs working rows, smokes one call, reports. Use when: When you need to absorb a GitHub repo into directory rows on the owner's voice command. Human article: /a/oip-capability-repo-absorb. Machine doc: ?key=REPO_ABSORB&format=markdown. Invocation history: /api/invocations?object_id=REPO_ABSORB.

SNAPSHOT_META

Repo snapshot metadata from KV: {sha, ts, byte_count} — no content blob. No args. Use this instead of REPO_SNAPSHOT whenever you only need to cite the sha, freshness, or size of the snapshot Use when: you need to snapshot meta Arguments: see content. Human article: /a/oip-capability-snapshot-meta. Machine doc: ?key=SNAPSHOT_META&format=markdown. Invocation history: /api/invocations?object_id=SNAPSHOT_META.

HISTORY_GET

How many past turns the build keeps per chat. No args. Returns {convo_max} Use when: the owner asks "how many messages do you remember" Arguments: none. Human article: /a/oip-capability-history-get. Machine doc: ?key=HISTORY_GET&format=markdown. Invocation history: /api/invocations?object_id=HISTORY_GET.

HISTORY_SET

Set how many past turns the build keeps per chat (1-100, default 14). Arg: the number. Takes effect immediately for every chat Use when: the owner says "remember more/fewer messages", "keep the last 30" Arguments: $1. Human article: /a/oip-capability-history-set. Machine doc: ?key=HISTORY_SET&format=markdown. Invocation history: /api/invocations?object_id=HISTORY_SET.

REASONING_GET

Current reasoning_effort for all xAI (grok) model calls. Returns {grok_reasoning_effort}. Values: low|medium|high|none|default. Default = model decides Use when: "what is the reasoning level set to" Arguments: none. Human article: /a/oip-capability-reasoning-get. Machine doc: ?key=REASONING_GET&format=markdown. Invocation history: /api/invocations?object_id=REASONING_GET.

REASONING_SET

Set reasoning_effort for all xAI (grok) model calls. Arg: low|medium|high|none|default. "default" removes the field (model decides). Takes effect immediately Use when: "set reasoning to low/high/off", "turn reasoning off", "set grok reasoning effort" Arguments: $1. Human article: /a/oip-capability-reasoning-set. Machine doc: ?key=REASONING_SET&format=markdown. Invocation history: /api/invocations?object_id=REASONING_SET.

AIG_LIST

List AI Gateways on the account Use when: you need to aig list Arguments: account_id. Human article: /a/oip-capability-aig-list. Machine doc: ?key=AIG_LIST&format=markdown. Invocation history: /api/invocations?object_id=AIG_LIST.

AIG_RAW

Raw call to AI Gateway REST. Arg $1 = full JSON body (model + messages) Use when: you need to aig raw Arguments: see content. Human article: /a/oip-capability-aig-raw. Machine doc: ?key=AIG_RAW&format=markdown. Invocation history: /api/invocations?object_id=AIG_RAW.

FIDELITY_REPORT

Per-run pass/fail counts from fidelity_log. Latest 5 runs. Use when: When you need to per-run pass/fail counts from fidelity_log. Human article: /a/oip-capability-fidelity-report. Machine doc: ?key=FIDELITY_REPORT&format=markdown. Invocation history: /api/invocations?object_id=FIDELITY_REPORT.

FIDELITY_RUN

Run the whole fidelity bank. $1 optional kind filter (positive | inverse | agent-route). Logs to fidelity_log. Returns JSON {run_id,total,passed,failed,duration_ms,failing} Use when: you need to fidelity run Arguments: $1. Human article: /a/oip-capability-fidelity-run. Machine doc: ?key=FIDELITY_RUN&format=markdown. Invocation history: /api/invocations?object_id=FIDELITY_RUN.

OPS

COMMERCIAL DATA — these tools exist. Use them for any revenue, order, customer or channel question. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-ops. Machine doc: ?key=OPS&format=markdown. Invocation history: /api/invocations?object_id=OPS.

HTTP_FETCH

Universal HTTP fetch. Method GET/POST/PUT/PATCH/DELETE. body is the raw request body string (empty for GET). headers_json is optional JSON header overrides. Returns 'HTTP <status>:<body first 20000 chars>'. Authorization redacted in ledger Use when: you need to http fetch Arguments: method|url|body|headers_json. Human article: /a/oip-capability-http-fetch. Machine doc: ?key=HTTP_FETCH&format=markdown. Invocation history: /api/invocations?object_id=HTTP_FETCH.

HTTP_ECHO_HEADERS

Echo the request headers this build sends to a provider, as httpbin sees them. A probe for connection account_selector (WT-0197): dispatch with with.account=httpbin:probe and the X-Account header appears in the answer. Use when: verifying that a connection's account_selector (header, query, host, path) reaches the wire. Arguments: none. Human article: /a/oip-capability-http-echo-headers. Machine doc: ?key=HTTP_ECHO_HEADERS&format=markdown. Invocation history: /api/invocations?object_id=HTTP_ECHO_HEADERS.

ART_GET

Read an article by slug Use when: you need to art get Arguments: see content. Human article: /a/oip-capability-art-get. Machine doc: ?key=ART_GET&format=markdown. Invocation history: /api/invocations?object_id=ART_GET.

ART_PATCH

Edit an article by slug Use when: you need to art patch Arguments: slug|json_body. Human article: /a/oip-capability-art-patch. Machine doc: ?key=ART_PATCH&format=markdown. Invocation history: /api/invocations?object_id=ART_PATCH.

WIRE_UP_AND_INVOKE

Atomic 3-step chain — resolve a short_id, append a wire-up cla Returns JSON {short_id,name,kind,appended_clause,append_result,invoke_key,invoke_result} Use when: ROUTER's prompt, and invoke the resolved tool Arguments: short_id|trigger_phrase|invoke_args. Human article: /a/oip-capability-wire-up-and-invoke. Machine doc: ?key=WIRE_UP_AND_INVOKE&format=markdown. Invocation history: /api/invocations?object_id=WIRE_UP_AND_INVOKE.

INVALIDATE_DIR_SNAPSHOT

Drop the KV cache of the directory snapshot. Use after writing to the directory table by hand or via any tool that does NOT already invalidate. Takes no args Use when: you need to invalidate dir snapshot Arguments: none. Human article: /a/oip-capability-invalidate-dir-snapshot. Machine doc: ?key=INVALIDATE_DIR_SNAPSHOT&format=markdown. Invocation history: /api/invocations?object_id=INVALIDATE_DIR_SNAPSHOT.

DISCOVER_SOURCE

Fetch a public URL (API docs, OpenAPI spec, GitHub raw, vendor SDK readme) and persist the raw body to R2 under capability_sources/<sha16>.txt. Returns {url, bytes, r2_key, sha256_16, status}. First step of universe-to-row. $1=URL Use when: you need to discover source Arguments: $1. Human article: /a/oip-capability-discover-source. Machine doc: ?key=DISCOVER_SOURCE&format=markdown. Invocation history: /api/invocations?object_id=DISCOVER_SOURCE.

EXTRACT_CAPABILITIES

Given an R2 key from DISCOVER_SOURCE, run a Workers AI Llama call to extract a JSON array of {op, method, url_or_signature, Truncates input at 20KB. Second step of universe-to-row. $1=r2_key, $2=optional model id (default @cf/meta/llama-3.3-70b-instruct-fp8-fast) Use when: you need to extract capabilities Arguments: {name,type,required}], description}. Human article: [/a/oip-capability-extract-capabilities. Machine doc: ?key=EXTRACT_CAPABILITIES&format=markdown. Invocation history: /api/invocations?object_id=EXTRACT_CAPABILITIES.

PROPOSE_ROWS

Given the JSON output of EXTRACT_CAPABILITIES (raw JSON string OR an r2_key starting with capability_sources/), emit SQL INSERTs for new directory rows, one per discovered op. Rows land disabled + planner_visible=0 so the owner can review before activating. $1=ops_json_or_r2_key, $2=category prefix (default "discovered") Use when: you need to propose rows Arguments: $1 | $2. Human article: /a/oip-capability-propose-rows. Machine doc: ?key=PROPOSE_ROWS&format=markdown. Invocation history: /api/invocations?object_id=PROPOSE_ROWS.

DELIVER_PENDING_ASSETS

Trigger the durable Workflow that drains pending_deliveries. Equivalent to the old /api/deliver heartbeat but with native retries and observability. Use when: When you need to trigger the durable Workflow that drains pending_deliveries. Human article: /a/oip-capability-deliver-pending-assets. Machine doc: ?key=DELIVER_PENDING_ASSETS&format=markdown. Invocation history: /api/invocations?object_id=DELIVER_PENDING_ASSETS.

WEBHOOK_INTAKE

Per-webhook intake (Blooio / 2chat / Telegram) collapsed to one flow. $1=channel (blooio|twochat|telegram), $2=sender, $3=text. Dedup, log, then route to BLOOIO_TURN_PHASE_A. Use when: When you need to per-webhook intake (Blooio / 2chat / Telegram) collapsed to one flow. Human article: /a/oip-capability-webhook-intake. Machine doc: ?key=WEBHOOK_INTAKE&format=markdown. Invocation history: /api/invocations?object_id=WEBHOOK_INTAKE.

DIRECT_EXEC_IF_PREFIX

If the inbound message starts with /t /exec /terminal /run /help, treat the rest as a LOCAL_EXEC command on the Mac bridge. Otherwise pass through unchanged. $1=full message. Use when: When you need to if the inbound message starts with /t /exec /terminal /run /help, treat the rest as a LOCAL_EXEC command on the Mac bridge. Human article: /a/oip-capability-direct-exec-if-prefix. Machine doc: ?key=DIRECT_EXEC_IF_PREFIX&format=markdown. Invocation history: /api/invocations?object_id=DIRECT_EXEC_IF_PREFIX.

MANIFEST

Self-describing directory: every callable row with description, runner, risk, requires_approval, status, input_schema, examples. The bootstrap contract for any client Use when: you need to manifest Arguments: category(optional). Human article: /a/oip-capability-manifest. Machine doc: ?key=MANIFEST&format=markdown. Invocation history: /api/invocations?object_id=MANIFEST.

SAVE_FLOW

Create or replace a flow row from DSL (pipe-safe) Use when: you need to save flow Arguments: key|dsl. Human article: /a/oip-capability-save-flow. Machine doc: ?key=SAVE_FLOW&format=markdown. Invocation history: /api/invocations?object_id=SAVE_FLOW.

SHORTCUT_RUN

catch-all for iOS Shortcuts that pass arbitrary JSON. Route to ROUTER. Use when: /api/phone/in action=shortcut_run. Arguments: payload_json.. Human article: /a/oip-capability-shortcut-run. Machine doc: ?key=SHORTCUT_RUN&format=markdown. Invocation history: /api/invocations?object_id=SHORTCUT_RUN.

PLANNER

Agent (model grok-4.3): PLANNER. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-planner. Machine doc: ?key=PLANNER&format=markdown. Invocation history: /api/invocations?object_id=PLANNER.

PROVIDERS

Provider/model registry — every company (xai|anthropic|openai|google), every model (text|image|video|stt|tts), with endpoint, api_key_name, every variable, cost in/out/cache, longest output, reasoning options, temperature range, and docs links. The on-hand API documentation behind the LLM-creation form Use when: you need to providers Arguments: see content. Human article: /a/oip-capability-providers. Machine doc: ?key=PROVIDERS&format=markdown. Invocation history: /api/invocations?object_id=PROVIDERS.

SITE_LINKS

Site surfaces (the only two human pages + every functional endpoint) Use when: you need to site links Arguments: see content. Human article: /a/oip-capability-site-links. Machine doc: ?key=SITE_LINKS&format=markdown. Invocation history: /api/invocations?object_id=SITE_LINKS.

PROVIDER_DOCS

LLM/model provider documentation (external). On-hand copies via PROVIDERS]<company>[/PROVIDERS] and [DOCS_GET]<slug>[/DOCS_GET] Use when: you need to provider docs Arguments: see content. Human article: [/a/oip-capability-provider-docs. Machine doc: ?key=PROVIDER_DOCS&format=markdown. Invocation history: /api/invocations?object_id=PROVIDER_DOCS.

PROVIDER_RECORD

A provider's standing as a read over the tables: claims by grade and state, methodologies used, verifications, disputes, agreements, counterparties, preregistered claims. Never a score. Use when: a buyer's AI decides whom to trust with what. Arguments: provider. Human article: /a/oip-capability-provider-record. Machine doc: ?key=PROVIDER_RECORD&format=markdown. Invocation history: /api/invocations?object_id=PROVIDER_RECORD.

TOOLING_DOCS

Platform + protocol references (external) Use when: you need to tooling docs Arguments: see content. Human article: /a/oip-capability-tooling-docs. Machine doc: ?key=TOOLING_DOCS&format=markdown. Invocation history: /api/invocations?object_id=TOOLING_DOCS.

CHANNEL_APIS

Channel + commerce API hosts (external). Tools that call them live as their own rows Use when: you need to channel apis Arguments: see content. Human article: /a/oip-capability-channel-apis. Machine doc: ?key=CHANNEL_APIS&format=markdown. Invocation history: /api/invocations?object_id=CHANNEL_APIS.

AUDIO

Speak words aloud. INVOKE: AUDIO]the words to speak[/AUDIO] Use when: you need to audio Arguments: $1 | $2. Human article: [/a/oip-capability-audio. Machine doc: ?key=AUDIO&format=markdown. Invocation history: /api/invocations?object_id=AUDIO.

RESCUE_ROUTER

Agent (model grok-4.3): RESCUE_ROUTER. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-rescue-router. Machine doc: ?key=RESCUE_ROUTER&format=markdown. Invocation history: /api/invocations?object_id=RESCUE_ROUTER.

DEMO_D1_TO_KV

demo — list D1 databases via the Cloudflare API, then store that JSON in KV. Proves sequential cross-surface composition. Use when: demonstration of combinatorial flows only. Human article: /a/oip-capability-demo-d1-to-kv. Machine doc: ?key=DEMO_D1_TO_KV&format=markdown. Invocation history: /api/invocations?object_id=DEMO_D1_TO_KV.

TOOLKIT

Agent (model grok-4.3): — You are TOOLKIT, a coding agent on the miscsubjects build with the full atomic toolkit: any Mac shell command, the named CLIs (wrangler/gh/npm/clasp), the Cloudflare REST API, the Google Workspace API, and the build registry itself. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-toolkit. Machine doc: ?key=TOOLKIT&format=markdown. Invocation history: /api/invocations?object_id=TOOLKIT.

CC_LAST

the last 3 Claude Code turns on this repo (the user input to Claude Code, the tools it used, the files it edited). Use when: "what did claude code just do", learning from Claude Code, the CC_MIRROR loop. Human article: /a/oip-capability-cc-last. Machine doc: ?key=CC_LAST&format=markdown. Invocation history: /api/invocations?object_id=CC_LAST.

CC_MIRROR

Agent (model grok-4.3): CC_MIRROR. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-cc-mirror. Machine doc: ?key=CC_MIRROR&format=markdown. Invocation history: /api/invocations?object_id=CC_MIRROR.

TODO_RUN

Run the next open task in the build to-do (tasks table) through the ROUTER. $1=mode; pass force to run on-demand. Cron calls it every 5 min, gated by KV todo_autorun Use when: you need to todo run Arguments: $1. Human article: /a/oip-capability-todo-run. Machine doc: ?key=TODO_RUN&format=markdown. Invocation history: /api/invocations?object_id=TODO_RUN.

COST_REPORT

Cost summary: total turns, total USD, average USD/turn, and the 10 priciest recent turns Use when: you need to cost report Arguments: none. Human article: /a/oip-capability-cost-report. Machine doc: ?key=COST_REPORT&format=markdown. Invocation history: /api/invocations?object_id=COST_REPORT.

CRITIC

Agent (model grok-4.3): CRITIC. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-critic. Machine doc: ?key=CRITIC&format=markdown. Invocation history: /api/invocations?object_id=CRITIC.

TEAM_RUN

Two agents reason against each other on a goal, bounded rounds, converge. $1=goal $2=proposer(default CODE_AUDIT) $3=critic(default CRITIC) $4=rounds(default 3) Use when: you need to team run Arguments: $1 | $2 | $3 | $4. Human article: /a/oip-capability-team-run. Machine doc: ?key=TEAM_RUN&format=markdown. Invocation history: /api/invocations?object_id=TEAM_RUN.

ROUTER

Agent (model grok-4.3): YOU ARE CURRENTLY PART OF AN API CALL IN WHICH YOUR ONLY OBJECTIVE IS TO UNDERSTAND THE INSTRUCTIONS YOU ARE NOW READING, HOW THEY RELATE TO THE TOOLS AVAILABLE TO YOU, AND TO ACT WITH REASONING THAT ANOTHER MODEL OR A HUMAN CAN AUDIT AFTERWARD FROM THE RECORD ALONE. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-router. Machine doc: ?key=ROUTER&format=markdown. Invocation history: /api/invocations?object_id=ROUTER.

CLOUDFLARE

Agent (model grok-4.3): the Cloudflare specialist in the owner's build. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-cloudflare. Machine doc: ?key=CLOUDFLARE&format=markdown. Invocation history: /api/invocations?object_id=CLOUDFLARE.

COMPUTER

Agent (model grok-4.3): the Computer specialist in the owner's build — you control his Mac. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-computer. Machine doc: ?key=COMPUTER&format=markdown. Invocation history: /api/invocations?object_id=COMPUTER.

DELTASK

delete a task. ARGS: <id>. EX {"key":"DELTASK","body":"42"} Human article: /a/oip-capability-deltask. Machine doc: ?key=DELTASK&format=markdown. Invocation history: /api/invocations?object_id=DELTASK.

EDITTASK

edit a task body. ARGS: <id>|<new text>. EX {"key":"EDITTASK","body":"42|buy milk"} Human article: /a/oip-capability-edittask. Machine doc: ?key=EDITTASK&format=markdown. Invocation history: /api/invocations?object_id=EDITTASK.

PEPTIDE_WRITER

PEPTIDE_WRITER — 57-article content map. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-peptide-writer. Machine doc: ?key=PEPTIDE_WRITER&format=markdown. Invocation history: /api/invocations?object_id=PEPTIDE_WRITER.

PEPTIDE_ROUTER

Before replying about a peptide, check if an internal article exists. $1=peptide name (e.g., bpc-157, tb-500, ara-290). Use when user asks about peptides or supplements. Use when: When you need to before replying about a peptide, check if an internal article exists. Human article: /a/oip-capability-peptide-router. Machine doc: ?key=PEPTIDE_ROUTER&format=markdown. Invocation history: /api/invocations?object_id=PEPTIDE_ROUTER.

REMEMBER

Additive memory. Append ONE line to a directory rows content without touching the rest. $1=row key (use ROUTER for your own memory), $2=the line. INVOKE: REMEMBER]ROUTER|- <one-line memory>[/REMEMBER] Use when: you need to remember Arguments: $1 | $2. Human article: [/a/oip-capability-remember. Machine doc: ?key=REMEMBER&format=markdown. Invocation history: /api/invocations?object_id=REMEMBER.

FORGET

Remove one line from your memory (or any row). $1=key (ROUTER for your own memory), $2=the exact text to remove Use when: you need to forget Arguments: $1 | $2. Human article: /a/oip-capability-forget. Machine doc: ?key=FORGET&format=markdown. Invocation history: /api/invocations?object_id=FORGET.

RESEARCH_BOT

Agent (model grok-4.3): RESEARCH_BOT, an agent that runs as a row in the directory and is dispatched by dispatch.js exactly like the ROUTER. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-research-bot. Machine doc: ?key=RESEARCH_BOT&format=markdown. Invocation history: /api/invocations?object_id=RESEARCH_BOT.

LOGIC

Agent (model grok-4.3): BE EXHAUSTIVE LOGICAL, DIRECT. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-logic. Machine doc: ?key=LOGIC&format=markdown. Invocation history: /api/invocations?object_id=LOGIC.

WRITER

Agent (model grok-4.3): Task: produce mechanism-led articles. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-writer. Machine doc: ?key=WRITER&format=markdown. Invocation history: /api/invocations?object_id=WRITER.

WRITER_AGENT

Agent (model grok-4.3): ARTICLE_WRITER. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-writer-agent. Machine doc: ?key=WRITER_AGENT&format=markdown. Invocation history: /api/invocations?object_id=WRITER_AGENT.

CARDS

Read the STATE CARDS. One card per isolated event: the inbound message (plain text), the reply (plain text), the agent system prompt, and every tool/CLI/file/http step with its raw payload in/out — each classified. Every card has a trace id (card_id) and a content hash, and is REST-callable Use when: you need to cards Arguments: see content. Human article: /a/oip-capability-cards. Machine doc: ?key=CARDS&format=markdown. Invocation history: /api/invocations?object_id=CARDS.

EVENTS

Read the raw EVENTS log — one row per call, newest first, chronological. Each row: source, key, action, status, request/response preview, trace id, step. $1 optional = text to search across key/action/payload Use when: you need to events Arguments: see content. Human article: /a/oip-capability-events. Machine doc: ?key=EVENTS&format=markdown. Invocation history: /api/invocations?object_id=EVENTS.

INTENT

Agent (model grok-4.3): you are a tool router. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-intent. Machine doc: ?key=INTENT&format=markdown. Invocation history: /api/invocations?object_id=INTENT.

PEPPER

Agent (model grok-4.3): you are Pepper, the peptide research assistant. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-pepper. Machine doc: ?key=PEPPER&format=markdown. Invocation history: /api/invocations?object_id=PEPPER.

GET_AGENT_LIMITS

Read current agent limits: tool_loops, memory_window, depth_cap, cost_cap_usd. Use when: the owner asks "what are my limits" / "what are your caps". Arguments: none. Human article: /a/oip-capability-get-agent-limits. Machine doc: ?key=GET_AGENT_LIMITS&format=markdown. Invocation history: /api/invocations?object_id=GET_AGENT_LIMITS.

NOW

Return the current time from the build clock in Pacific time (America/Los_Angeles). Use when: When you need to return the current time from the build clock in Pacific time (America/Los_Angeles). Arguments: none. Human article: /a/oip-capability-now. Machine doc: ?key=NOW&format=markdown. Invocation history: /api/invocations?object_id=NOW.

CRON_LIST

List existing cron jobs by querying the Cloudflare API. Returns scheduled jobs for this account. Use when: When you need to list existing cron jobs by querying the Cloudflare API. Human article: /a/oip-capability-cron-list. Machine doc: ?key=CRON_LIST&format=markdown. Invocation history: /api/invocations?object_id=CRON_LIST.

MEMORY_AGENT

Agent (model grok-4.3): the MEMORY_AGENT. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-memory-agent. Machine doc: ?key=MEMORY_AGENT&format=markdown. Invocation history: /api/invocations?object_id=MEMORY_AGENT.

EDITOR_AGENT

Agent (model grok-4.3): ARTICLE_EDITOR. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-editor-agent. Machine doc: ?key=EDITOR_AGENT&format=markdown. Invocation history: /api/invocations?object_id=EDITOR_AGENT.

POPULATE

Grok+web-search adds evidence sources to a peptide article (2 rounds). Use when: When you need to grok+web-search adds evidence sources to a peptide article (2 rounds). Human article: /a/oip-capability-populate. Machine doc: ?key=POPULATE&format=markdown. Invocation history: /api/invocations?object_id=POPULATE.

IDEAS

Grok+web-search returns ideas, publishes nothing. Use when: When you need to grok+web-search returns ideas, publishes nothing. Human article: /a/oip-capability-ideas. Machine doc: ?key=IDEAS&format=markdown. Invocation history: /api/invocations?object_id=IDEAS.

WRITE_ARTICLE

Grok writes + publishes an article. Use when: When you need to grok writes + publishes an article. Human article: /a/oip-capability-write-article. Machine doc: ?key=WRITE_ARTICLE&format=markdown. Invocation history: /api/invocations?object_id=WRITE_ARTICLE.

BLOOIO2

Agent (model grok-4.3): the owner's writing assistant on the second Blooio line (+12065711028). Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-blooio2. Machine doc: ?key=BLOOIO2&format=markdown. Invocation history: /api/invocations?object_id=BLOOIO2.

ADDTASK

Add a job to the build queue (tasks table). Cron will pick it up and run it automatically when protocol_autorun is on.\n# WHEN_TO_USE: the owner wants to batch work and not wait for a live reply.\n# ARGS: = JSON job object OR plain text ask. JSON keys: ask (required), role (default writer), slug, topic, web_search (default true), max_tokens, post_to (default /api/protocol/write), model.\n# EX: ADDTASK]{"ask":"Write evidence-graded BPC-157 article","slug":"bpc-157","web_search":true}[/ADDTASK]\n# EX: [ADDTASK]{"ask":"Populate BPC-157 with more sources and widgets","slug":"bpc-157","post_to":"/api/ Use when: the owner wants to batch work and not wait for a live reply.\n# ARGS: = JSON job object OR plain text ask. JSON keys: ask (required), role (default writer), slug, topic, web_search (default true), max_tokens, post_to (default /api/protocol/write), model.\n# EX: [ADDTASK]{"ask":"Write evidence-graded BPC-157 article","slug":"bpc-157","web_search":true}[/ADDTASK]\n# EX: [ADDTASK]{"ask":"Populate BPC-157 with more sources and widgets","slug":"bpc-157","post_to":"/api/protocol/populate","max_rounds":4}[/ADDTASK]\n["","ADDTASK"] Arguments: = JSON job object OR plain text ask. JSON keys: ask (required), role (default writer), slug, topic, web_search (default true), max_tokens, post_to (default /api/protocol/write), model.\n# EX: [ADDTASK]{"ask":"Write evidence-graded BPC-157 article","slug":"bpc-157","web_search":true}[/ADDTASK]\n# EX: [ADDTASK]{"ask":"Populate BPC-157 with more sources and widgets","slug":"bpc-157","post_to":"/api/protocol/populate","max_rounds":4}[/ADDTASK]\n["","ADDTASK"]. Human article: [/a/oip-capability-addtask. Machine doc: ?key=ADDTASK&format=markdown. Invocation history: /api/invocations?object_id=ADDTASK.

REGEX_PARSE

Invoke fn regexParse with positional args "$1"]. Use when: When you need to invoke fn regexParse with positional args ["$1"]. Human article: [/a/oip-capability-regex-parse. Machine doc: ?key=REGEX_PARSE&format=markdown. Invocation history: /api/invocations?object_id=REGEX_PARSE.

DONETASK

Mark a task as done. $1=task_id (number), $2=optional result summary. Use when: a queued task has been completed and should be removed from the open backlog. Arguments: $1=task_id, $2=result summary. Human article: /a/oip-capability-donetask. Machine doc: ?key=DONETASK&format=markdown. Invocation history: /api/invocations?object_id=DONETASK.

CHARLIE_PROSECUTOR

Charlie Prosecutor. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-charlie-prosecutor. Machine doc: ?key=CHARLIE_PROSECUTOR&format=markdown. Invocation history: /api/invocations?object_id=CHARLIE_PROSECUTOR.

CHARLIE_CRITIC

Charlie Critic. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-charlie-critic. Machine doc: ?key=CHARLIE_CRITIC&format=markdown. Invocation history: /api/invocations?object_id=CHARLIE_CRITIC.

TIME_NOW

Return the current UTC ISO timestamp.\n# EX: TIME_NOW][/TIME_NOW]\n[] Human article: [/a/oip-capability-time-now. Machine doc: ?key=TIME_NOW&format=markdown. Invocation history: /api/invocations?object_id=TIME_NOW.

PROACTIVE_POKE

Send one short proactive text to the owner. Used by the cron; you can also trigger manually.\n# EX: PROACTIVE_POKE][/PROACTIVE_POKE]\n[] Human article: [/a/oip-capability-proactive-poke. Machine doc: ?key=PROACTIVE_POKE&format=markdown. Invocation history: /api/invocations?object_id=PROACTIVE_POKE.

SCHEDULERS

List scheduler/autorun state: KV loop flags plus configured automations. Use when: owner asks what loops, schedulers, cron, autoruns, or recurring jobs are on/off. Arguments: none. Human article: /a/oip-capability-schedulers. Machine doc: ?key=SCHEDULERS&format=markdown. Invocation history: /api/invocations?object_id=SCHEDULERS.

SCHEDULER_SET

Enable or disable a cron scheduler. $1=key, $2=value (0 or 1). Human article: /a/oip-capability-scheduler-set. Machine doc: ?key=SCHEDULER_SET&format=markdown. Invocation history: /api/invocations?object_id=SCHEDULER_SET.

KNOWLEDGE

KNOWLEDGE — fetch a knowledge card for models. Use when: When you need to kNOWLEDGE — fetch a knowledge card for models. Arguments: CUSTOMER_FUNNEL | AD_ACCOUNTS | MARKETING_STATE | LEO_RESEARCH | IMAGE_REFERENCE | REACTIONS | BATCH | ARTICLE_LIST. Human article: /a/oip-capability-knowledge. Machine doc: ?key=KNOWLEDGE&format=markdown. Invocation history: /api/invocations?object_id=KNOWLEDGE.

REFLEX_PASS

Live probes vs protocol vision claims — posts reflex conformance atoms with proves/responds_to edges. Use when: after collaborate/audit, continuous self-proof loop, verify build topology matches vision. Arguments: slug (optional, default protocol). Human article: /a/oip-capability-reflex-pass. Machine doc: ?key=REFLEX_PASS&format=markdown. Invocation history: /api/invocations?object_id=REFLEX_PASS.

GRAPH_GROW

Run one model growth queue tick — populate sources, Kimi, Gemini, repair, reflex. Use when: cron, after deploy, proactive graph enrichment. Arguments: optional slug|step e.g. bpc-157|kimi_collaborate. Human article: /a/oip-capability-graph-grow. Machine doc: ?key=GRAPH_GROW&format=markdown. Invocation history: /api/invocations?object_id=GRAPH_GROW.

GRAPH_LINT

Graph maintenance pass over the whole corpus — orphans (no inbound links), missing pages (wikilinked but unwritten), unsourced claims, open challenges, stale hub pages. Use when: after any publish, on cron, before picking new work, when asked "what is wrong with the graph". Arguments: none. Human article: /a/oip-capability-graph-lint. Machine doc: ?key=GRAPH_LINT&format=markdown. Invocation history: /api/invocations?object_id=GRAPH_LINT.

OBSIDIAN_PULL

Pull Obsidian vault export to local folder (via LOCAL_EXEC). Arguments: optional slugs comma-separated. Human article: /a/oip-capability-obsidian-pull. Machine doc: ?key=OBSIDIAN_PULL&format=markdown. Invocation history: /api/invocations?object_id=OBSIDIAN_PULL.

PLAYWRIGHT_START

Start the Playwright MCP server on the Mac bridge\n# WHEN_TO_USE: before calling PLAYWRIGHT if the server is not running\n# ARGS: none\n# EX: PLAYWRIGHT_START][/PLAYWRIGHT_START]\n{"cmd":"sh","args":["-lc","/Users/owner/miscsubjects-pages/bridge/playwright-mcp-launcher.sh"],"timeout":600000} Use when: before calling PLAYWRIGHT if the server is not running\n# ARGS: none\n# EX: [PLAYWRIGHT_START][/PLAYWRIGHT_START]\n{"cmd":"sh","args":["-lc","/Users/owner/miscsubjects-pages/bridge/playwright-mcp-launcher.sh"],"timeout":600000} Arguments: none\n# EX: [PLAYWRIGHT_START][/PLAYWRIGHT_START]\n{"cmd":"sh","args":["-lc","/Users/owner/miscsubjects-pages/bridge/playwright-mcp-launcher.sh"],"timeout":600000}. Human article: [/a/oip-capability-playwright-start. Machine doc: ?key=PLAYWRIGHT_START&format=markdown. Invocation history: /api/invocations?object_id=PLAYWRIGHT_START.

PLAYWRIGHT

Proxy a tool call to the Playwright MCP server on the Mac bridge Use when: drive a real browser: navigate, click, type, screenshot, evaluate JS, manage tabs Arguments: $1=tool_name, $2+=JSON arguments. Human article: /a/oip-capability-playwright. Machine doc: ?key=PLAYWRIGHT&format=markdown. Invocation history: /api/invocations?object_id=PLAYWRIGHT.

NOTIFY_OWNER

Text the owner (the owner) a message — the "background daemon pings you when it needs a decision" primitive. Automations and event hooks call this to ask a question or report. Use when: the AI needs the owner's input or wants to surface something — "tell the owner", "ask the owner", "ping me when X". Arguments: message. Human article: /a/oip-capability-notify-owner. Machine doc: ?key=NOTIFY_OWNER&format=markdown. Invocation history: /api/invocations?object_id=NOTIFY_OWNER.

OPEN_URL

Open a URL in the default browser on the owner's Mac (a new tab). No Accessibility needed. Use when: "open a browser", "open this website", "pull up <site>", "search YouTube/Google for X" (pass the search URL). Arguments: the URL (a page or a search URL like https://www.youtube.com/results?search_query=lofi). Human article: /a/oip-capability-open-url. Machine doc: ?key=OPEN_URL&format=markdown. Invocation history: /api/invocations?object_id=OPEN_URL.

MODEL_CHAT_INTAKE

Append raw outside-model/chat text to the ledger and queue the receiving editorial board. Use when: paste any model answer, raw chat log, critique, complaint, or documentation feedback into the build so the board extracts rules and queues purification. Arguments: $1+ raw text/plain chat log. Human article: /a/oip-capability-model-chat-intake. Machine doc: ?key=MODEL_CHAT_INTAKE&format=markdown. Invocation history: /api/invocations?object_id=MODEL_CHAT_INTAKE.

EDITORIAL_BOARD_RUN

Run one receiving editorial-board task. It reads a MODEL_CHAT_INTAKE ledger event, extracts owner complaints and content-rule defects as JSON, ledgers EDITORIAL_BOARD_DECISION, and queues OIP purification. Use when: after raw model/chat intake, or cron, to process one editorial-board queue item. Arguments: none. Human article: /a/oip-capability-editorial-board-run. Machine doc: ?key=EDITORIAL_BOARD_RUN&format=markdown. Invocation history: /api/invocations?object_id=EDITORIAL_BOARD_RUN.

TRY_STRIPE_MCP

Try the Stripe integration before connecting it. Shows the read and write objects Stripe exposes here and recommends connect or skip. Financial: the live read-only account check runs only in mode "live". Use when: "should I get Stripe MCP", "what would Stripe let an agent do". Arguments: $1 = optional mode "live". Human article: /a/oip-capability-try-stripe-mcp. Machine doc: ?key=TRY_STRIPE_MCP&format=markdown. Invocation history: /api/invocations?object_id=TRY_STRIPE_MCP.

TRY_GITHUB_MCP

Try the GitHub integration before connecting it. Runs a safe read-only trial (list issues) and returns a receipt. Use when: "should I get GitHub MCP", "what would GitHub let an agent do". Arguments: none. Human article: /a/oip-capability-try-github-mcp. Machine doc: ?key=TRY_GITHUB_MCP&format=markdown. Invocation history: /api/invocations?object_id=TRY_GITHUB_MCP.

ISSUE_SWEEP

Autonomous issue closer. Scans open issues and closes safe categories. Use when: run a batch of issue closures; schedule via AUTOMATE_ADD for recursive loops. Arguments: mode (triage | docs | all). Human article: /a/oip-capability-issue-sweep. Machine doc: ?key=ISSUE_SWEEP&format=markdown. Invocation history: /api/invocations?object_id=ISSUE_SWEEP.

QUADSYNC_RUN

Run the server half of QUADSYNC now — mirror new ledger events to GitHub (ledger-mirror/events-<day>.jsonl) and fold recent GitHub commits + auto] issues back into the ledger/tasks. Returns both results plus all four corner health stamps. Use when: the owner says "sync", "sync everything", "run quadsync", "is everything synced" — or any model needs the corners current before reasoning about build state. Automatic every 10 min via dispatch traffic; local Mac + Google Drive corners run via launchd com.the owner.miscsubjects.quadsync. Arguments: none. Human article: [/a/oip-capability-quadsync-run. Machine doc: ?key=QUADSYNC_RUN&format=markdown. Invocation history: /api/invocations?object_id=QUADSYNC_RUN.

STALE_ISSUE_DIGEST

Return a digest of open GitHub issues that have not been updated in N days. Use when: weekly purge review, stale backlog triage. Arguments: days. Human article: /a/oip-capability-stale-issue-digest. Machine doc: ?key=STALE_ISSUE_DIGEST&format=markdown. Invocation history: /api/invocations?object_id=STALE_ISSUE_DIGEST.

OBJECTION_LOG

File an objection, confirm a duplicate, settle an exact objection, or append a repair without erasing the original. Arguments: one JSON object. New: {slug,body,claimed_model,target_div?,stance?}. Duplicate confirmation: add duplicate_of:"obj-N". Repair/answer lane: add repairs:"obj-N" (or answer_of), body describing the correction and answer or stance:"upgrade". The repair bypasses similarity rejection, preserves the original, and appends linked discourse.. Human article: /a/oip-capability-objection-log. Machine doc: ?key=OBJECTION_LOG&format=markdown. Invocation history: /api/invocations?object_id=OBJECTION_LOG.

OBJECTION_JUDGE_V1

Agent (model grok-4.3): OBJECTION JUDGE — v1. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-objection-judge-v1. Machine doc: ?key=OBJECTION_JUDGE_V1&format=markdown. Invocation history: /api/invocations?object_id=OBJECTION_JUDGE_V1.

PROSECUTOR_RUN

One machine turn of the operator loop, end to end: fetch the drop + current accepted thread-state, ask a model for ONE materially new point (inheriting all accepted state, never repeating it), and post the result to the thread bus as a proposed update. Replies NOTHING NEW when the state already covers everything it sees. Use when: the owner says "prosecute the protocol", "run the loop", "have a machine critique it" — or the governor wants fresh adversarial load without any human transport. Arguments: model key (optional; default ASK_CLAUDE — also ASK_GPT / ASK_GEMINI / ASK_KIMI). Human article: /a/oip-capability-prosecutor-run. Machine doc: ?key=PROSECUTOR_RUN&format=markdown. Invocation history: /api/invocations?object_id=PROSECUTOR_RUN.

EAGLE_IMSG

Eagle image review workflow from iMessage. Runs eagle_imessage.mjs on the Mac. Texts the owner image previews + generated variations. Updates EAGLE_IMAGES sheet. Use when: the owner asks about eagles, eagle images, show eagles, generate eagle variation, mark eagle good, eagle prompts — any eagle1–eagle25 image work over iMessage. Arguments: his message (pass through verbatim). Human article: /a/oip-capability-eagle-imsg. Machine doc: ?key=EAGLE_IMSG&format=markdown. Invocation history: /api/invocations?object_id=EAGLE_IMSG.

MARKETING_SNAPSHOT

Live Meta + lbl cache + MARKETING_STATE JSON. Human article: /a/oip-capability-marketing-snapshot. Machine doc: ?key=MARKETING_SNAPSHOT&format=markdown. Invocation history: /api/invocations?object_id=MARKETING_SNAPSHOT.

UI_SURFACE_PROBE

Compare operator-visible fetch (no terminal key) vs agent fetch — ledgered mismatch flag. Use when: Before claiming any admin page or live URL works; after deploy of user-visible UI. Arguments: $1=url path or full URL; optional $2=extra|markers|pipe|delimited. Human article: /a/oip-capability-ui-surface-probe. Machine doc: ?key=UI_SURFACE_PROBE&format=markdown. Invocation history: /api/invocations?object_id=UI_SURFACE_PROBE.

OUTSTANDING_SYNC_EMAIL

Build one unified outstanding-sync report and email it: every open GitHub issue, open/running task inventory, Kimi CLI/Desktop sync state, and historical recurring problem classes. Use when: The owner asks to sync GitHub issues, tasks, Kimi, model work, outstanding work, or wants everything sent by email. Arguments: mode. blank/force = send email + short iMessage notice. dry = collect only, no send.. Human article: /a/oip-capability-outstanding-sync-email. Machine doc: ?key=OUTSTANDING_SYNC_EMAIL&format=markdown. Invocation history: /api/invocations?object_id=OUTSTANDING_SYNC_EMAIL.

WORK_FEED

The shared work thread. THE FIRST THING ANY MODEL OR TOKEN-HOLDER READS. Aggregates the top priority, open tasks, open GitHub issues, the lead pipeline, and the recent model thread into one ranked view. Use when: at the start of ANY session, or when asked 'what should I work on', 'what is the state', 'catch me up', 'where are we'. Arguments: $1 = how many recent thread entries (default 15).. Human article: /a/oip-capability-work-feed. Machine doc: ?key=WORK_FEED&format=markdown. Invocation history: /api/invocations?object_id=WORK_FEED.

WORK_APPEND

Append one entry to the shared, append-only work thread. Models here have short memories and there are many of them; this thread is how the next one continues from what you did instead of rediscovering it. Nothing in the thread is ever edited or deleted — a correction is a new entry that names what it supersedes. Use when: After you do, propose, or audit anything. Always append — a change nobody recorded is a change the next agent will undo. Arguments: model_agent (required) — Who is writing, e · kind (required) — What kind of entry this is · entry (required) — What happened, in plain sentences. Human article: /a/oip-capability-work-append. Machine doc: ?key=WORK_APPEND&format=markdown. Invocation history: /api/invocations?object_id=WORK_APPEND.

ARXIV_PAPER

The arXiv paper as a live object. The paper "The Document Is the Receipt" lives at github.com/[OWNER_HANDLE]/oip (private) and is written only by ARXIV_GROW. Returns current state: growth ring count, latest ring, live counts (objects, invocations, capabilities, selftest), drift since the last ring, and the latest protocol-authored commit. Use when: the owner asks "paper state", "how big is the paper", "when did the paper last grow", "show the arxiv object", "has the paper drifted". Arguments: none.. Human article: /a/oip-capability-arxiv-paper. Machine doc: ?key=ARXIV_PAPER&format=markdown. Invocation history: /api/invocations?object_id=ARXIV_PAPER.

ARXIV_GROW

Regenerate the arXiv paper from live state. Reads paper/template.tex + paper/rings.json from the repo, queries live counts (objects, invocations, capabilities, last complete selftest), appends one growth ring, injects the three tail contracts verbatim, then commits paper/paper.tex + paper/rings.json + README.md + oip.json — each commit message carries this trace id. CI compiles the PDF on the paper.tex push. This fn is the only writer of the generated files. Use when: the owner says "grow the paper", "regenerate the arxiv", "add a ring", "refresh the paper". Also fired daily by launchd com.the owner.oip.arxiv-grow on the Mac. Arguments: none.. Human article: /a/oip-capability-arxiv-grow. Machine doc: ?key=ARXIV_GROW&format=markdown. Invocation history: /api/invocations?object_id=ARXIV_GROW.

AUDIT_PROBE_ROW

Audit probe row - safe to delete. Use when: When you need to audit probe row - safe to delete. Human article: /a/oip-capability-audit-probe-row. Machine doc: ?key=AUDIT_PROBE_ROW&format=markdown. Invocation history: /api/invocations?object_id=AUDIT_PROBE_ROW.

AUDIT_COLD_CASE_V1

Agent (model grok-4.3): Numbered clauses; a lower number beats a higher one on conflict. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-audit-cold-case-v1. Machine doc: ?key=AUDIT_COLD_CASE_V1&format=markdown. Invocation history: /api/invocations?object_id=AUDIT_COLD_CASE_V1.

FED_SEND

Send a signed federation message (oip-message/1) from the home agent to a remote agent at ANOTHER domain, and return that agent's signed reply. Cross-domain agent-to-agent messaging; the two current test nodes are separately deployed under one operator. Use when: coordinate with an agent on a different domain. The recipient's signing key + inbox come from its https://<domain>/.well-known/oip.json; the reply signature is verified before it is trusted. Arguments: recipient|kind|rest — recipient=agent@domain. kind=query|invoke.. Human article: /a/oip-capability-fed-send. Machine doc: ?key=FED_SEND&format=markdown. Invocation history: /api/invocations?object_id=FED_SEND.

RELAY_POST_APPEND

Append one model's public adoption/proof link to THE RELAY. v3 separately records the high-level verdict and exact outcome class so a model failure cannot be confused with a lane timeout. Use when: After a model audits the prior relay and performs real work. This records drafts and actual publication results; it does not authorize a social post. Arguments: one JSON object: platform; identity_mode named|incognito; exact model_name, model_provider, model_version and session_label; action; result_summary; verdict PASS|FAIL|MIXED; outcome_class SUCCESS|PARTIAL|MODEL_FAILED|LANE_TIMEOUT (PASS=SUCCESS, MIXED=PARTIAL, FAIL uses one of the two failure classes); proof_links]; media_links[]; platform_copy with LinkedIn/Facebook/Instagram/X required, each beginning [execution surface · exact model name · YYYY-MM-DD HH:MM UTC] then a newline and third-person observed result; tag_targets[{name,handle?,why}] with at least one materially connected target; pub. Human article: [/a/oip-capability-relay-post-append. Machine doc: ?key=RELAY_POST_APPEND&format=markdown. Invocation history: /api/invocations?object_id=RELAY_POST_APPEND.

DEPLOY_LEASE

Inspect, acquire or release the single production deployment door for loop-safe-miscsubjects. The canonical ship script holds the same KV lease from before migrations through the Pages result and ledgers acquire/release. Arguments: op check|acquire|release | holder | nonce. Acquire returns a 30-minute nonce. Release requires the exact nonce. Check is read-only.. Human article: /a/oip-capability-deploy-lease. Machine doc: ?key=DEPLOY_LEASE&format=markdown. Invocation history: /api/invocations?object_id=DEPLOY_LEASE.

DISCLOSURE_GET

Read a versioned public defensive-publication artifact from the disclosure archive. Text is scanned for bearer/credential material at read time; binary artifacts are admitted only after local render/hash/credential review. Keys are immutable and public. Arguments: $1 = public disclosure path returned by a publication manifest, for example 2026-07-17/operation-killbox-v1.1/specification.md.. Human article: /a/oip-capability-disclosure-get. Machine doc: ?key=DISCLOSURE_GET&format=markdown. Invocation history: /api/invocations?object_id=DISCLOSURE_GET.

STANDARD_REGISTER

Register a versioned standard whose clauses can be cited by decision records. This records the source and authority class; it does not turn advisory text into law. Arguments: JSON {id,name,version,authority_class:internal-profile|external-source|advisory|legal-review-required,source_url?,canonical_text,clauses:{id,title,requirement,test?,authority?}],status?,parent_id?,created_by}.. Human article: [/a/oip-capability-standard-register. Machine doc: ?key=STANDARD_REGISTER&format=markdown. Invocation history: /api/invocations?object_id=STANDARD_REGISTER.

REVIEW_RECORD

Confirm, challenge or abstain on a decision record while preserving reviewer provider/family, evidence, prompt/context fingerprints and whether prior answers were visible. Arguments: JSON {decision_id,reviewer_model,reviewer_provider,reviewer_family,stance:CONFIRM|CHALLENGE|ABSTAIN,justification,evidence],evidence_recomputed?,prompt_hash?,context_hash?,prior_answers_visible?,authority,invocation_id?}.. Human article: [/a/oip-capability-review-record. Machine doc: ?key=REVIEW_RECORD&format=markdown. Invocation history: /api/invocations?object_id=REVIEW_RECORD.

SURETY_RECORD

Compute the disclosed independence-weighted support/challenge profile for one decision. Surety measures corroboration, not truth, legality or consensus authority. Arguments: JSON {decision_id}.. Human article: /a/oip-capability-surety-record. Machine doc: ?key=SURETY_RECORD&format=markdown. Invocation history: /api/invocations?object_id=SURETY_RECORD.

CERTIFIER_HISTORY

Read the cards, revocations, expiries and evidence history filed by a named regulator, insurer, auditor, compliance officer, standards body or owner. Arguments: JSON {certifier_label}.. Human article: /a/oip-capability-certifier-history. Machine doc: ?key=CERTIFIER_HISTORY&format=markdown. Invocation history: /api/invocations?object_id=CERTIFIER_HISTORY.

PRIVACY_EGRESS

Shape or record context egress under the OIP Privacy Profile. Dry-run classification/minimization never executes a transmission; authorization binds an exact hash/recipient/purpose; disclosure and erasure outcomes remain distinct. Arguments: JSON {action:shape|classify|minimize|authorize|disclose|erasure|conformance,...}. Read https://miscsubjects.com/api/privacy for exact per-action schemas. For real sensitive payloads call the owner-authenticated endpoint directly; dispatch receipts store invocation input, so use this row only with synthetic or already-redacted content.. Human article: /a/oip-capability-privacy-egress. Machine doc: ?key=PRIVACY_EGRESS&format=markdown. Invocation history: /api/invocations?object_id=PRIVACY_EGRESS.

CITATION_VALIDATION

Independently validate that one cited evidence item actually supports the clause finding it was filed under. A model confirming a decision is NOT citation validation; this records source existence, version/hash correctness, passage-to-premise support, clause-to-conduct applicability, material omissions and conclusion overreach, plus the honest evidence class. Arguments: JSON {decision_id,clause,evidence_ref,evidence_class:operator-served|independently-recomputable|third-party-witnessed|institutionally-attested|private-scoped|unresolved-assertion,verdict:SUPPORTED|PARTIALLY_SUPPORTED|UNSUPPORTED|CONTRADICTED|LEGAL_REVIEW_REQUIRED,source_exists?,version_hash_correct?,passage_supports_premise?,clause_governs_conduct?,material_omission?,conclusion_overreach?,validator_model,validator_provider,validator_family,prompt_hash?,context_hash?,prior_answers_visible?,recompute_method?,justification}.. Human article: /a/oip-capability-citation-validation. Machine doc: ?key=CITATION_VALIDATION&format=markdown. Invocation history: /api/invocations?object_id=CITATION_VALIDATION.

COMPLIANCE_GATE

Ask a bounded compliance card to authorize a consequential operation. Proves the card is executable state: a currently valid, in-scope, correct-version, in-jurisdiction, within-risk, dissent-clear, correctly-certified card permits; anything else returns a typed, receipted denial. Uses a safe demonstration operation and never gates production-critical behavior. Arguments: JSON {card_id,requested_action,system_version?,jurisdiction?,risk?,required_certifier_type?,presented_card_hash?,require_no_standing_dissent?,actor?}.. Human article: /a/oip-capability-compliance-gate. Machine doc: ?key=COMPLIANCE_GATE&format=markdown. Invocation history: /api/invocations?object_id=COMPLIANCE_GATE.

DESIGN_LAW

Discover the canonical Laws of Design Knowledge-Action Object.\n# WHEN_TO_USE: any design, rendering, navigation, editorial, graph, admin, source, SEO, readability, or article-skill decision.\n# ARGS: none for the contract; follow representations.invoke with surface and task to apply it.\n# EX: DESIGN_LAW][/DESIGN_LAW]\n# CANONICAL: https://miscsubjects.com/api/articles/design-law\n# NOTE: this row is the discovery facet; meaning, instructions, failures, tests, versions, graph, and representations live in the canonical object and are not duplicated here. Use when: any design, rendering, navigation, editorial, graph, admin, source, SEO, readability, or article-skill decision.\n# ARGS: none for the contract; follow representations.invoke with surface and task to apply it.\n# EX: [DESIGN_LAW][/DESIGN_LAW]\n# CANONICAL: https://miscsubjects.com/api/articles/design-law\n# NOTE: this row is the discovery facet; meaning, instructions, failures, tests, versions, graph, and representations live in the canonical object and are not duplicated here. Arguments: none for the contract; follow representations.invoke with surface and task to apply it.\n# EX: [DESIGN_LAW][/DESIGN_LAW]\n# CANONICAL: https://miscsubjects.com/api/articles/design-law\n# NOTE: this row is the discovery facet; meaning, instructions, failures, tests, versions, graph, and representations live in the canonical object and are not duplicated here.. Human article: [/a/oip-capability-design-law. Machine doc: ?key=DESIGN_LAW&format=markdown. Invocation history: /api/invocations?object_id=DESIGN_LAW.

OP_ROOT

Read OP, the Object Protocol: definition, invariants, canonical roots, and OIP compatibility boundary. Arguments: None. Add ?format=markdown for a model-readable document.. Human article: /a/oip-capability-op-root. Machine doc: ?key=OP_ROOT&format=markdown. Invocation history: /api/invocations?object_id=OP_ROOT.

TAP_GO_MODEL_PROFILES

Read the five owner-editable model-specific content slots used by token Tap & Go: ChatGPT, Claude, Grok, Gemini, and Kimi. The model selector belongs to the token DROP, not the build audit. Arguments: None for read. Owner edits one profile with PUT /api/tap-go-profiles {model,content}.. Human article: /a/oip-capability-tap-go-model-profiles. Machine doc: ?key=TAP_GO_MODEL_PROFILES&format=markdown. Invocation history: /api/invocations?object_id=TAP_GO_MODEL_PROFILES.

LANDSCAPE_NEXT

Return one queued external-build research object with the common axes and the existing voxel-batch source/claim append shape.\n# ARGS: none.\n# EX: LANDSCAPE_NEXT][/LANDSCAPE_NEXT]\n[] Arguments: none.\n# EX: [LANDSCAPE_NEXT][/LANDSCAPE_NEXT]\n[]. Human article: [/a/oip-capability-landscape-next. Machine doc: ?key=LANDSCAPE_NEXT&format=markdown. Invocation history: /api/invocations?object_id=LANDSCAPE_NEXT.

NORMANDY_ASSIGNMENT

Read one reserved outside-model contribution slot: current graph snapshot, named comparison target, shared axis, already-stored limits, additive slots, and the existing voxel-batch write lane.\n# ARGS: $1=assignment id from a minted build-audit DROP.\n# EX: NORMANDY_ASSIGNMENT]norm-abc123[/NORMANDY_ASSIGNMENT]\n["$1"] Arguments: $1=assignment id from a minted build-audit DROP.\n# EX: [NORMANDY_ASSIGNMENT]norm-abc123[/NORMANDY_ASSIGNMENT]\n["$1"]. Human article: [/a/oip-capability-normandy-assignment. Machine doc: ?key=NORMANDY_ASSIGNMENT&format=markdown. Invocation history: /api/invocations?object_id=NORMANDY_ASSIGNMENT.

CONTACT_GET

Get a lead as a canonical CONTACT object (channels, fit score, signals, provenance). Arguments: $1 = lead id or contact:<id>.. Human article: /a/oip-capability-contact-get. Machine doc: ?key=CONTACT_GET&format=markdown. Invocation history: /api/invocations?object_id=CONTACT_GET.

CONTACTS_LIST

List contact objects with optional filter. Arguments: $1 = JSON {status?,segment?,city?,limit?} or a bare status word.. Human article: /a/oip-capability-contacts-list. Machine doc: ?key=CONTACTS_LIST&format=markdown. Invocation history: /api/invocations?object_id=CONTACTS_LIST.

WRITING_LAW

Discover the canonical Laws of Writing Knowledge-Action Object — opaque, decorative, ambiguous language is hostility toward the reader.\n# WHEN_TO_USE: any writing, rewriting, editing, replying, header, summary, argument, concession, or challenge decision on any surface.\n# ARGS: none for the contract; follow representations.invoke with text and task to apply it.\n# EX: WRITING_LAW][/WRITING_LAW]\n# CANONICAL: https://miscsubjects.com/api/articles/writing-law\n# NOTE: this row is the discovery facet; meaning, instructions, failures, tests, versions, graph, and representations live in the cano Use when: any writing, rewriting, editing, replying, header, summary, argument, concession, or challenge decision on any surface.\n# ARGS: none for the contract; follow representations.invoke with text and task to apply it.\n# EX: [WRITING_LAW][/WRITING_LAW]\n# CANONICAL: https://miscsubjects.com/api/articles/writing-law\n# NOTE: this row is the discovery facet; meaning, instructions, failures, tests, versions, graph, and representations live in the canonical object and are not duplicated here. Arguments: none for the contract; follow representations.invoke with text and task to apply it.\n# EX: [WRITING_LAW][/WRITING_LAW]\n# CANONICAL: https://miscsubjects.com/api/articles/writing-law\n# NOTE: this row is the discovery facet; meaning, instructions, failures, tests, versions, graph, and representations live in the canonical object and are not duplicated here.. Human article: [/a/oip-capability-writing-law. Machine doc: ?key=WRITING_LAW&format=markdown. Invocation history: /api/invocations?object_id=WRITING_LAW.

PROOF_PING

GET the /api/proof page and return the raw HTML for verification. Use when: the user wants to check the proof/liveness endpoint. Human article: /a/oip-capability-proof-ping. Machine doc: ?key=PROOF_PING&format=markdown. Invocation history: /api/invocations?object_id=PROOF_PING.

WIKIPEDIA_SUMMARY

Fetch the canonical Wikipedia REST summary for one page title (extract, description, canonical URL, revision id) for use as a citable source in an article. Use when: sourcing a claim that needs an encyclopedic reference with a stable revision id; pair with the article webhook kind=source. Arguments: $1 = page title, underscores for spaces.. Human article: /a/oip-capability-wikipedia-summary. Machine doc: ?key=WIKIPEDIA_SUMMARY&format=markdown. Invocation history: /api/invocations?object_id=WIKIPEDIA_SUMMARY.

WITNESS_MINT

Mint a WITNESS token: read-only authority over ONE adjudication, bound to a named audience, revocable, with its own ledger trail. Three parties can each hold one over the same finding; none holds operator authority and none must trust the others. A token forwarded to any party other than its audience fails closed. Every use is recorded. Use when: any finding more than one party must check independently. Proves independent VERIFICATION, not independent execution. Arguments: $1 = adjudication id (inv_...) · $2 = audience the token is bound to · $3 = ttl seconds (use 604800 for 7 days). Human article: /a/oip-capability-witness-mint. Machine doc: ?key=WITNESS_MINT&format=markdown. Invocation history: /api/invocations?object_id=WITNESS_MINT.

SEAL_PANEL

The sealer. Deterministic arithmetic over a panel's findings that decides what happens to the ACTION and nothing else. No model runs at this position: a model here is a further opinion that can share the panel's blind spot while being the thing that decides. Five outcomes, all arithmetic: APPROVE (unanimous AFFIRM, identical clause citations, enough distinct training families, no malformed finding), NEGATE (unanimous DENY on the same terms - the action is refused, not deferred), NO_ACTION (unanimous CANNOT_CONCLUDE - a required record is missing, so nothing is authorised and nothing is refused Use when: at the end of every panel whose finding will reach a downstream actor. Clause-citation divergence fires before verdict divergence and is the more sensitive detector, so run this rather than counting votes. Arguments: one JSON object {findings:{model,verdict,clauses|reasoning,confidence?,invocation_id,exposure,role}], min_families?, min_findings?, min_confidence?, escalate_to?}. Human article: [/a/oip-capability-seal-panel. Machine doc: ?key=SEAL_PANEL&format=markdown. Invocation history: /api/invocations?object_id=SEAL_PANEL.

ALLOCATE_REASONING

The runtime allocator. Turns an action and its action class into R (loss exposure), K (complexity) and epsilon (permitted wrongful-action rate) from a VERSIONED SERVER-OWNED policy, selects the least-cost configuration whose MEASURED undetected-wrong rate is at or below that epsilon, executes it so every model payload lands on the ledger, seals it with SEAL_PANEL bound to those records, and then performs the bounded downstream act only if the seal returns APPROVE. NEGATE refuses the act. NO_ACTION leaves it untouched. DISPUTE and ESCALATE create a human-review object bound to a NAMED reviewer Use when: any consequential action that must not execute until enough auditable reasoning has been purchased for its consequence. Arguments: one JSON object {action, action_class, question, ruleset_url, ruleset_hash, rules], artifact, artifact_hash, task_class?, reviewer?, reviewer_audience?}. The caller does NOT supply R, K, epsilon, thresholds or the configuration.. Human article: [/a/oip-capability-allocate-reasoning. Machine doc: ?key=ALLOCATE_REASONING&format=markdown. Invocation history: /api/invocations?object_id=ALLOCATE_REASONING.

PROMO_CLASSES

Read the audience-class table — the starting logic of the self-promotion loop, as data. Each row: the loss the class bears, the mechanism here that reduces it, the strongest artifact to show them, the reply-earning thesis, their first objection, fit 0-100 with reason, prior, last contact. Use when: before any allocation, draft, or outreach plan for the build itself; "who would care about this". Arguments: $1 = class key (optional; empty = all, fit desc).. Human article: /a/oip-capability-promo-classes. Machine doc: ?key=PROMO_CLASSES&format=markdown. Invocation history: /api/invocations?object_id=PROMO_CLASSES.

CONSCIENCE_GATE

The Good Conscience Law — the veto between "can execute" and "will execute". MAY_ACT = authority AND evidence AND conscience; logical economics optimizes only among MAY_ACT=true actions. Empty body returns the constitution (build-conscience@1.0.0, clauses GC1-GC8). A REFUSE/ESCALATE/HALT verdict is rejected unless it names the violated clause, the prohibited consequence, the job's direct causal contribution, and evidence — refusal binds to a named clause, never to free moralizing. HALT writes KV conscience:halt: every outbound category (email, leads, x, reddit, messaging, self-promotion) refus Use when: before the build accepts any job or takes any consequential outbound action; when work smells like it violates the floor; "should the build do this at all". Arguments: $1 = empty (list clauses) OR JSON {job, verdict:ACCEPT|REFUSE|ESCALATE|HALT, violated_clause?, prohibited_consequence?, causal_contribution?, evidence?, notes?}. Human article: /a/oip-capability-conscience-gate. Machine doc: ?key=CONSCIENCE_GATE&format=markdown. Invocation history: /api/invocations?object_id=CONSCIENCE_GATE.

BOARD_TICK

run the owner sheet. Executes every APPROVE / REJECT / SAVE / PULL / RUN the owner typed in Arguments: none. WHEN_TO_USE: on a schedule, so what the owner types in the sheet happens by itself.. Human article: /a/oip-capability-board-tick. Machine doc: ?key=BOARD_TICK&format=markdown. Invocation history: /api/invocations?object_id=BOARD_TICK.

CONSTITUTION_V1_4_0

Agent (model grok-4.3): THE DECISION CONSTITUTION — decision-constitution@1.4.0 Human article: /a/oip-capability-constitution-v1-4-0. Machine doc: ?key=CONSTITUTION_V1_4_0&format=markdown. Invocation history: /api/invocations?object_id=CONSTITUTION_V1_4_0.

PROVEN_WORK_SPEC

PROVEN_WORK_SPEC — the evolving product spec every loop rep must meet. Versioned; amend by appending a version block with the exhibit (a field-audit criticism that survived). Canonical definition: https://miscsubjects.com/a/proven-work. Use when: When you need to pROVEN_WORK_SPEC — the evolving product spec every loop rep must meet. Human article: /a/oip-capability-proven-work-spec. Machine doc: ?key=PROVEN_WORK_SPEC&format=markdown. Invocation history: /api/invocations?object_id=PROVEN_WORK_SPEC.

NEXT_ACTS

The ranked "what should be written next" queue, derived live from the graph — write (missing wikilinked pages), resolve (challenged claims), source (unsourced claims), revise (stale hubs), connect (orphans), respond (unread replies), outreach (quiet high-fit classes). Use when: at the start of any content/outreach session, on cron, when asked "what next". Arguments: optional limit (default 10, max 50). Human article: /a/oip-capability-next-acts. Machine doc: ?key=NEXT_ACTS&format=markdown. Invocation history: /api/invocations?object_id=NEXT_ACTS.

OWNER_REPORT

Send a report to the owner and prove it arrived. Returns the inbound ledger row id that witnessed the message crossing the internet, or an explicit not_witnessed failure. Never reports a send on faith. Use when: every report, inventory, digest or answer addressed to the owner. Use EMAIL_SEND_TRACKED for outreach to third parties instead. Arguments: $1 = subject (the OWNER REPORT] prefix is added). $2 = body text. $3 = seconds to wait for the ledger row (default 30, max 60; observed round trip is about 25s).. Human article: [/a/oip-capability-owner-report. Machine doc: ?key=OWNER_REPORT&format=markdown. Invocation history: /api/invocations?object_id=OWNER_REPORT.

CLOCK_TICK

Fire every automation anchored to a wall-clock hour whose hour has arrived, once per local day. Anchored rows (trigger=clock, at_hour set) never fire on an interval — this is the only path that runs them. Use when: the every-few-minutes tick calls it; call it by hand to see what is anchored and what is waiting. Arguments: none. Human article: /a/oip-capability-clock-tick. Machine doc: ?key=CLOCK_TICK&format=markdown. Invocation history: /api/invocations?object_id=CLOCK_TICK.

WT0090_OUTBOUND_AUDIT_V1

Agent (model grok-4.3): Lower-numbered clauses win on conflict. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-wt0090-outbound-audit-v1. Machine doc: ?key=WT0090_OUTBOUND_AUDIT_V1&format=markdown. Invocation history: /api/invocations?object_id=WT0090_OUTBOUND_AUDIT_V1.

WT0090_OUTBOUND_POLISH_V1

Agent (model @cf/meta/llama-3.3-70b-instruct-fp8-fast): Write one exact first-contact email and its companion X post from the INPUT row. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-wt0090-outbound-polish-v1. Machine doc: ?key=WT0090_OUTBOUND_POLISH_V1&format=markdown. Invocation history: /api/invocations?object_id=WT0090_OUTBOUND_POLISH_V1.

REASON

Agent (model grok-4.3): YOU ARE CURRENTLY PART OF AN API CALL IN WHICH YOUR ONLY OBJECTIVE IS TO UNDERSTAND THE INSTRUCTIONS YOU ARE NOW READING, HOW THEY RELATE TO THE TOOLS AVAILABLE TO YOU, AND TO ACT WITH REASONING THAT ANOTHER MODEL OR A HUMAN CAN AUDIT AFTERWARD FROM THE RECORD ALONE. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-reason. Machine doc: ?key=REASON&format=markdown. Invocation history: /api/invocations?object_id=REASON.

CHAIN_SEAL

Fold newly-logged ledger events into the public transparency chain: appends a checkpoint, builds the batch's Merkle tree, and signs the checkpoint with the home ES256 key. Caps at 10,000 leaves per call, so a backlog needs repeated calls. Use when: on a schedule, and any time /api/chain reports an event_count far below SELECT COUNT(*) FROM events. An unsealed event is not yet covered by a published head, so it is not yet tamper-evident to an outsider. Arguments: none. Human article: /a/oip-capability-chain-seal. Machine doc: ?key=CHAIN_SEAL&format=markdown. Invocation history: /api/invocations?object_id=CHAIN_SEAL.

GORGIAS_TICKETS

Support and recovery tickets, newest first. $1 = how many (max 100). Use when: "what are customers complaining about", "recovery outreach", churn signals. Arguments: $1 = limit. Human article: /a/oip-capability-gorgias-tickets. Machine doc: ?key=GORGIAS_TICKETS&format=markdown. Invocation history: /api/invocations?object_id=GORGIAS_TICKETS.

RESEND_EMAILS

Transactional email sends with delivery state. $1 = how many (max 100). Use when: "did the order confirmation go out", bounce and complaint checks. Arguments: $1 = limit. Human article: /a/oip-capability-resend-emails. Machine doc: ?key=RESEND_EMAILS&format=markdown. Invocation history: /api/invocations?object_id=RESEND_EMAILS.

BASH

Run a shell command on the owner Mac (alias of LOCAL_EXEC; agents write BASH]cmd[/BASH]). Use when: any file operation, git command, system check, or script execution. Arguments: $1+ = full shell command (pipes/OR/chains preserved).. Human article: [/a/oip-capability-bash. Machine doc: ?key=BASH&format=markdown. Invocation history: /api/invocations?object_id=BASH.

ENVIRONMENT

Canonical, recursively self-describing environment root. Human article: /a/oip-capability-environment. Machine doc: ?key=ENVIRONMENT&format=markdown. Invocation history: /api/invocations?object_id=ENVIRONMENT.

EXTERNAL_GOOGLE_SHEETS

External comparable for mutable grid plus documented scripting grammar. Human article: /a/oip-capability-external-google-sheets. Machine doc: ?key=EXTERNAL_GOOGLE_SHEETS&format=markdown. Invocation history: /api/invocations?object_id=EXTERNAL_GOOGLE_SHEETS.

CHATGPT_WEB

Ask ChatGPT Web — the logged-in browser session, not the metered API — and return the exact captured answer. Use when: you want ChatGPT Web specifically as a step: a second opinion, a council member, or a reviewer of another capability's output. Arguments: $1 = the prompt. Pipes are preserved.. Human article: /a/oip-capability-chatgpt-web. Machine doc: ?key=CHATGPT_WEB&format=markdown. Invocation history: /api/invocations?object_id=CHATGPT_WEB.

CLAUDE_WEB

Ask Claude Web — the logged-in browser session, not the metered API — and return the exact captured answer. Use when: you want Claude Web specifically as a step: a second opinion, a council member, or a reviewer of another capability's output. Arguments: $1 = the prompt. Pipes are preserved.. Human article: /a/oip-capability-claude-web. Machine doc: ?key=CLAUDE_WEB&format=markdown. Invocation history: /api/invocations?object_id=CLAUDE_WEB.

RIGHTS_SET

Record the owner's right to sell one directory row: the entitlement behind it, whether resale is allowed, whether the row is read-only. A row with no rights record cannot be quoted, linked or granted. Use when: before any row is sold or rented; the owner names what makes it sellable. Arguments: row_key|entitlement_source|resale_allowed(1/0)|read_only(1/0)|platform_terms_version|expires_at|purpose|region. Human article: /a/oip-capability-rights-set. Machine doc: ?key=RIGHTS_SET&format=markdown. Invocation history: /api/invocations?object_id=RIGHTS_SET.

RIGHTS_GET

Read the rights record of one row. Use when: checking whether a row may be sold and on what entitlement. Arguments: row_key. Human article: /a/oip-capability-rights-get. Machine doc: ?key=RIGHTS_GET&format=markdown. Invocation history: /api/invocations?object_id=RIGHTS_GET.

QUOTE

Price one request: marginal cost for the units, the settlement fee, a risk premium and collateral when a scarce resource is involved. A quote is never one number hiding another. Use when: a buyer or their AI asks what a row or a rental will cost before paying. Arguments: row_key|units|resource_id. Human article: /a/oip-capability-quote. Machine doc: ?key=QUOTE&format=markdown. Invocation history: /api/invocations?object_id=QUOTE.

PAY_LINK

Create the Stripe product, price and payment link for one priced row with a rights record. The payment collects the buyer's phone; the webhook mints the token. A Stripe write: the owner's go-ahead phrase must be in the third field. Use when: the owner opens a row for sale. Arguments: row_key|units|go ahead and create the payment link. Human article: /a/oip-capability-pay-link. Machine doc: ?key=PAY_LINK&format=markdown. Invocation history: /api/invocations?object_id=PAY_LINK.

PAID_GRANT

The pay-then-token step: mint a capability scoped to the bought row for the bought units, bind it to the payer's profile by phone or email, deliver it by text, record the sale and its settlement terms. Called by the Stripe webhook after a verified payment, or by the owner for a comped grant. Never twice for one payment. Use when: a payment was verified, or the owner comps a buyer. Arguments: row_key|units|buyer_phone|buyer_email|source_id|amount_cents. Human article: /a/oip-capability-paid-grant. Machine doc: ?key=PAID_GRANT&format=markdown. Invocation history: /api/invocations?object_id=PAID_GRANT.

PAID_LANE_LOOKUP

What a phone number may run: the live paid capabilities bound to its profile with uses left and expiry. Use when: the messaging pre-route decides whether an inbound text belongs to a paying sender. Arguments: phone. Human article: /a/oip-capability-paid-lane-lookup. Machine doc: ?key=PAID_LANE_LOOKUP&format=markdown. Invocation history: /api/invocations?object_id=PAID_LANE_LOOKUP.

RESOURCE_NEW

Declare a scarce resource with occupancy and state: an advertising account, a logged-in browser profile, a seat, a compute slot. Names the capability and fields that snapshot its material state, and its collateral and risk premium. Use when: the owner offers something for lease rather than per-call use. Arguments: kind|label|snapshot_key|snapshot_body|snapshot_fields(csv)|collateral_usd|risk_premium_usd. Human article: /a/oip-capability-resource-new. Machine doc: ?key=RESOURCE_NEW&format=markdown. Invocation history: /api/invocations?object_id=RESOURCE_NEW.

RESOURCE_SNAPSHOT

Hash the material state of a resource now: the fields it declares, read through its snapshot capability. Use when: before and after a lease, or to check a resource by hand. Arguments: resource_id|lease_id|phase. Human article: /a/oip-capability-resource-snapshot. Machine doc: ?key=RESOURCE_SNAPSHOT&format=markdown. Invocation history: /api/invocations?object_id=RESOURCE_SNAPSHOT.

MANDATE_SET

What the buyer's AI may do about a want: budget, deadline, provider classes, allowed and forbidden actions, the price under which it may execute without asking, the price above which the person must confirm, the evidence required, cancellation terms. Nothing executes without one. Use when: right after a want, before any solicitation. Arguments: want_id|budget_usd|deadline|provider_classes|allowed_actions|forbidden_actions|auto_execute_usd|confirm_usd|evidence_required|cancellation. Human article: /a/oip-capability-mandate-set. Machine doc: ?key=MANDATE_SET&format=markdown. Invocation history: /api/invocations?object_id=MANDATE_SET.

OFFER_NEW

A provider answers a solicitation: scope, method, price, timing, evidence promised, and whether the work is human, AI or hybrid, with the grade of authority behind the offer. Use when: a provider or their AI bids on a solicitation. Arguments: solicitation_id|provider|scope|method|price_usd|timing|evidence_promised|composition(human/ai/hybrid)|authority_grade. Human article: /a/oip-capability-offer-new. Machine doc: ?key=OFFER_NEW&format=markdown. Invocation history: /api/invocations?object_id=OFFER_NEW.

AGREEMENT_ACCEPT

The buyer accepts one offer under the mandate; obligations become binding; a work task is created with the acceptance tests and a settlement is held in escrow until the tests pass. Above the auto-execute threshold the person must approve. Use when: the buyer chooses an offer. Arguments: offer_id|acceptance_tests_json|authority_grade. Human article: /a/oip-capability-agreement-accept. Machine doc: ?key=AGREEMENT_ACCEPT&format=markdown. Invocation history: /api/invocations?object_id=AGREEMENT_ACCEPT.

METHODOLOGY

The versioned definition a claim family is computed by: statement, required evidence, exclusions, grade ceiling. Without a key, lists every methodology. Use when: reading how a claim is computed before making or verifying one. Arguments: key. Human article: /a/oip-capability-methodology. Machine doc: ?key=METHODOLOGY&format=markdown. Invocation history: /api/invocations?object_id=METHODOLOGY.

NODE_RECORD

This node's record: its signing key (public half), domain, agent id, subscribed governance facets and the kernel hash it accepted. record writes it (owner), read returns it. Use when: a federation peer or a reader asks who this node is and which facets it honours. Arguments: read, or record|facets(csv)|note. Human article: /a/oip-capability-node-record. Machine doc: ?key=NODE_RECORD&format=markdown. Invocation history: /api/invocations?object_id=NODE_RECORD.

OFFERS_RANK

Rank the open offers on a solicitation into a set under the buyer's trust policy: best verified, cheapest acceptable, fastest acceptable, and the offers excluded by the policy. Never one winner; payment never moves an offer into the trusted set. Use when: the buyer's AI compares offers. Arguments: solicitation_id. Human article: /a/oip-capability-offers-rank. Machine doc: ?key=OFFERS_RANK&format=markdown. Invocation history: /api/invocations?object_id=OFFERS_RANK.

TRUST_POLICY_SET

A buyer's trust policy: the verifier set, accepted methods, minimum evidence grade, maximum evidence age, conflict and appeal policy. Replaces one global score. Use when: a buyer decides whose claims count for them. Arguments: profile_id|minimum_grade|max_evidence_age_days|verifier_set(csv)|accepted_methods(csv)|conflict_policy|appeal_policy. Human article: /a/oip-capability-trust-policy-set. Machine doc: ?key=TRUST_POLICY_SET&format=markdown. Invocation history: /api/invocations?object_id=TRUST_POLICY_SET.

TRUST_POLICY_GET

Read a profile's trust policy. Use when: checking how a buyer will judge claims and offers. Arguments: profile_id. Human article: /a/oip-capability-trust-policy-get. Machine doc: ?key=TRUST_POLICY_GET&format=markdown. Invocation history: /api/invocations?object_id=TRUST_POLICY_GET.

PANIC_REVOKE

End every token bound to a profile, device or session at once, end its leases, revoke its devices. One call. Use when: something leaked. Arguments: profile or device or session|id|reason. Human article: /a/oip-capability-panic-revoke. Machine doc: ?key=PANIC_REVOKE&format=markdown. Invocation history: /api/invocations?object_id=PANIC_REVOKE.

AGGREGATE_QUERY

An aggregate over a declared dataset: sum, mean, median, min, max over at least five members, never the rows. Each buyer has a budget of distinct queries per dataset; a cohort that differs from an earlier answered one by fewer than five members is refused, because the difference would isolate them. Use when: a buyer wants a benchmark without anyone's rows. Arguments: buyer|dataset(sales_by_row, claims_by_method, leases_by_kind)|filters_json. Human article: /a/oip-capability-aggregate-query. Machine doc: ?key=AGGREGATE_QUERY&format=markdown. Invocation history: /api/invocations?object_id=AGGREGATE_QUERY.

HOLD_RECORD

Record a bond on a challenge or an escrow hold on an agreement. The card authorization is a Stripe write and moves to the card only with the owner's phrase; without it the hold is recorded and named as such. Use when: a challenger stakes a bond; an agreement wants real escrow. Arguments: bond or escrow|ref_id|party|amount_usd|go ahead and hold. Human article: /a/oip-capability-hold-record. Machine doc: ?key=HOLD_RECORD&format=markdown. Invocation history: /api/invocations?object_id=HOLD_RECORD.

CPC_DELTA_V1

Methodology CPC_DELTA_V1, version 1: percentage change in cost per click between a baseline window and a test window, from two Meta insights receipts. Rows with zero clicks are excluded; a window with zero clicks makes the claim EVIDENCE_MISSING. Grade ceiling: method-reproducible. Use when: a claim about a change in cost per click. Arguments: none. Human article: /a/oip-capability-cpc-delta-v1. Machine doc: ?key=CPC_DELTA_V1&format=markdown. Invocation history: /api/invocations?object_id=CPC_DELTA_V1.

CPC_LIFT_V1

Methodology CPC_LIFT_V1, version 1: difference-in-differences lift in cost per click over a test cohort and a control cohort, four Meta insights receipts. Grade ceiling: causally-supported when the commitment named the control before the windows were read. Use when: a claim that something caused a change in cost per click. Arguments: none. Human article: /a/oip-capability-cpc-lift-v1. Machine doc: ?key=CPC_LIFT_V1&format=markdown. Invocation history: /api/invocations?object_id=CPC_LIFT_V1.

ROAS_V1

Methodology ROAS_V1, version 1: revenue divided by ad spend over one window, from a Meta insights receipt and a Stripe charges receipt; refunded charges excluded; attribution of revenue to spend is not claimed. Grade ceiling: method-reproducible. Use when: a claim about return on ad spend. Arguments: none. Human article: /a/oip-capability-roas-v1. Machine doc: ?key=ROAS_V1&format=markdown. Invocation history: /api/invocations?object_id=ROAS_V1.

POLICY_STEP_UP_ACCOUNT_CHANGES

A policy row read at the one authorization door: any token invoking a Meta account-changing row (budget, status, delete) must present a device with a human verification inside the last ten minutes. Policies tighten bindings and never loosen them; disable this row and the requirement lifts with no deploy. Use when: read by contextGateCheck on every invocation; not run by hand. Arguments: none (declarative). Human article: /a/oip-capability-policy-step-up-account-changes. Machine doc: ?key=POLICY_STEP_UP_ACCOUNT_CHANGES&format=markdown. Invocation history: /api/invocations?object_id=POLICY_STEP_UP_ACCOUNT_CHANGES.

CANONICAL_MANUAL

The canonical invocation manual as markdown: how to do anything on this build by every surface (POST, GET with a token, tags, MCP, sheet cell, browser model), with the exact recorded request and response for each. Use when: a model or person with zero context needs to know how to call, prove, delegate, compose or publish here. Read it before guessing any call. Arguments: none. Human article: /a/oip-capability-canonical-manual. Machine doc: ?key=CANONICAL_MANUAL&format=markdown. Invocation history: /api/invocations?object_id=CANONICAL_MANUAL.

WEEKEND_ERROR_DIGEST

On a weekend, text the owner the last thing that errored. On a weekday it does nothing at all. Use when: the same job comes up again. Call this instead of doing the steps by hand; it composes into other flows, schedules and every projection like any row. Arguments: $1 = the argument the first step takes.. Human article: /a/oip-capability-weekend-error-digest. Machine doc: ?key=WEEKEND_ERROR_DIGEST&format=markdown. Invocation history: /api/invocations?object_id=WEEKEND_ERROR_DIGEST.

PULSE_THREE_AT_ONCE

Check the clock, and if it answered, ask three sources at the same time. Use when: the same job comes up again. Call this instead of doing the steps by hand; it composes into other flows, schedules and every projection like any row. Arguments: $1 = the argument the first step takes.. Human article: /a/oip-capability-pulse-three-at-once. Machine doc: ?key=PULSE_THREE_AT_ONCE&format=markdown. Invocation history: /api/invocations?object_id=PULSE_THREE_AT_ONCE.

AD_LIBRARY_SCRAPE

Collect public Meta Ad Library results via system Chrome. Graph ads_archive is locked (error 10); this is the public transparency page. Saves every ad to ~/.miscsubjects/ad-library/YYYY-MM-DD/ads.jsonl. Use when: competitor ads, peptide terms, volume search without the restricted API. Arguments: empty = all queries in scripts/ad-library-queries.json. Or --q TERM.. Human article: /a/oip-capability-ad-library-scrape. Machine doc: ?key=AD_LIBRARY_SCRAPE&format=markdown. Invocation history: /api/invocations?object_id=AD_LIBRARY_SCRAPE.

APPROVE_DEMO

The pause, end to end. Reads the clock, stops in front of the send, and only sends once a person says yes. Use when: to see APPROVE and FLOW_RESUME work, and as the shape to copy for any flow that touches the world. Arguments: $1 = note=<the text to send>. Named, with a default, so the order does not matter.. Human article: /a/oip-capability-approve-demo. Machine doc: ?key=APPROVE_DEMO&format=markdown. Invocation history: /api/invocations?object_id=APPROVE_DEMO.

INLINE_GATE_PROBE

Probe: a gate on one step, and a named parameter with a default. Use when: to see that a false inline gate skips its own step and the flow carries on. Arguments: $1 = word=go (default: skip). Human article: /a/oip-capability-inline-gate-probe. Machine doc: ?key=INLINE_GATE_PROBE&format=markdown. Invocation history: /api/invocations?object_id=INLINE_GATE_PROBE.

GTASKS_LISTS

The connected account's Google Tasks lists. Use when: find a list id before reading its tasks. Arguments: none. Human article: /a/oip-capability-gtasks-lists. Machine doc: ?key=GTASKS_LISTS&format=markdown. Invocation history: /api/invocations?object_id=GTASKS_LISTS.

GTASKS_ITEMS

Open tasks in one Google Tasks list. Use when: what is on the to-do list. @default is the primary list. Arguments: $1 = list id or @default. Human article: /a/oip-capability-gtasks-items. Machine doc: ?key=GTASKS_ITEMS&format=markdown. Invocation history: /api/invocations?object_id=GTASKS_ITEMS.

GWORKSPACE_SEARCH

Search across Gmail, Drive, Docs, Sheets, Calendar and Chat at once (Workspace Universal Search MCP). Use when: "where did I see X" across Google Workspace. Arguments: $1+ = query. Human article: /a/oip-capability-gworkspace-search. Machine doc: ?key=GWORKSPACE_SEARCH&format=markdown. Invocation history: /api/invocations?object_id=GWORKSPACE_SEARCH.

GDOCS_MCP_READ

Read a Google Doc structure and text (Docs MCP). Use when: after GDRIVE_MCP_SEARCH finds a doc. Arguments: $1 = document id. Human article: /a/oip-capability-gdocs-mcp-read. Machine doc: ?key=GDOCS_MCP_READ&format=markdown. Invocation history: /api/invocations?object_id=GDOCS_MCP_READ.

GCHAT_MCP_SEARCH

Search Google Chat messages the owner can see (Chat MCP) by keywords. Use when: find a Chat message by words. Arguments: $1+ = keywords. Human article: /a/oip-capability-gchat-mcp-search. Machine doc: ?key=GCHAT_MCP_SEARCH&format=markdown. Invocation history: /api/invocations?object_id=GCHAT_MCP_SEARCH.

RESOLVE

Turn one line of the invocation grammar (or a plain-English request) into a routed, receipted capability call. Resolves key, args, account, device and authority in one step; runs it when run=true. Use when: you have words, not a key: "quake feed 3 5", "message send +1415… hi", "screenshot example.com". Also the door every CLI, SDK and function-calling runtime uses. Arguments: $1+ = the line. Human article: /a/oip-capability-resolve. Machine doc: ?key=RESOLVE&format=markdown. Invocation history: /api/invocations?object_id=RESOLVE.

FIND

Find capabilities by use case. Ranked over key words, WHAT and WHEN_TO_USE text, aliases and examples; returns each hit with its argument schema and its CLI, tool and SDK forms. Use when: before guessing a key. "find text someone", "find screenshot", "find stripe refund". Arguments: $1+ = what you want to do. Human article: /a/oip-capability-find. Machine doc: ?key=FIND&format=markdown. Invocation history: /api/invocations?object_id=FIND.

SUBJECT_LINK_ADD

Link a person (traffic profile) to an object: receipt://inv_…, work://WT-…, external://stripe/customer/cus_…, sheet://…, article://… with a relation (created_by, acted_on, attributed_to, came_from, converted_to, member_of, observed_on, assigned_to, belongs_to). The identity spine reaching the objects. Use when: "this customer is that Stripe customer", attributing a receipt or a task to a person. Arguments: $1 = profile id (prf_…), $2 = object ref, $3 = relation, $4 = evidence source (owner_assertion | provider_match | capability_context | cookie, default owner_assertion). Human article: /a/oip-capability-subject-link-add. Machine doc: ?key=SUBJECT_LINK_ADD&format=markdown. Invocation history: /api/invocations?object_id=SUBJECT_LINK_ADD.

SUBJECT_LINKS

The links of a person or an object: give a profile id (prf_…) to see everything linked to that person, or an object ref (receipt://…, external://…) to see who it belongs to. Use when: "what do we know about this person", "whose receipt is this". Arguments: $1 = profile id or object ref. Human article: /a/oip-capability-subject-links. Machine doc: ?key=SUBJECT_LINKS&format=markdown. Invocation history: /api/invocations?object_id=SUBJECT_LINKS.

CREDS

The measured map of every credential this build holds - vault names, already-authenticated CLIs, key files, and the named gaps with the one action that closes each. Never prints a value. Use when: BEFORE asking the owner for any token, key, password or login. Before saying a credential is missing. When you need to know where something lives. Arguments: $1 = '' (human table) | --json | --gaps. Human article: /a/oip-capability-creds. Machine doc: ?key=CREDS&format=markdown. Invocation history: /api/invocations?object_id=CREDS.

DS_OR

Agent (model deepseek/deepseek-v4-flash-0731): DeepSeek V4 Flash via OpenRouter. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-ds-or. Machine doc: ?key=DS_OR&format=markdown. Invocation history: /api/invocations?object_id=DS_OR.

DS_CF

Agent (model openrouter/deepseek/deepseek-v4-flash-0731): DeepSeek V4 Flash via Cloudflare AI Gateway. Use when: When the dispatcher routes a turn to this agent, or you want this voice/model on a task. Human article: /a/oip-capability-ds-cf. Machine doc: ?key=DS_CF&format=markdown. Invocation history: /api/invocations?object_id=DS_CF.

CODEBASE_SEARCH

Semantic search over this repo's indexed source (functions/, scripts/, workers/, apps-script/, migrations/, schema.sql) via the CODEBASE_INDEX Vectorize index. Returns matched chunks with path, start_line and the chunk's own text — no second call needed to read the file. $1=query, $2=optional top_k (default 8, max 25) Use when: a coding agent needs to find the relevant file(s) for a task instead of holding the whole repo in context Arguments: $1 | $2. Human article: /a/oip-capability-codebase-search. Machine doc: ?key=CODEBASE_SEARCH&format=markdown. Invocation history: /api/invocations?object_id=CODEBASE_SEARCH.

TURN_JOB_RUN

Internal. Runs one agent turn that the Blooio path queued (functions/blooio.js hopToAgent), inside a request the TASKS queue consumer holds open, so an agent whose tools are slow (Moby) is not cut off after thirty seconds. The job is read from turn_jobs and runs once. Use when: never by hand. The queue sends it. Arguments: $1 = turn_jobs id, $2 = the job's one-time token. A wrong token is refused.. Human article: /a/oip-capability-turn-job-run. Machine doc: ?key=TURN_JOB_RUN&format=markdown. Invocation history: /api/invocations?object_id=TURN_JOB_RUN.

EXA_SEARCH

Search the web with Exa and get back the best pages with the passages that matched, rather than a page of links. Neural search, so plain description works better than keywords. Use when: anything current, any claim that needs a source, finding companies or people or documents by description. Arguments: $1 = what to search for | $2 = how many results (default 10) | $3 = type: auto|fast|instant|deep-lite|deep|deep-reasoning (default auto). Human article: /a/oip-capability-exa-search. Machine doc: ?key=EXA_SEARCH&format=markdown. Invocation history: /api/invocations?object_id=EXA_SEARCH.

ENVS_LIST

The environment registry (console_environments): each Worker or lane a model can be sent work in — what it is, where it runs, which Cloudflare products are attached, what it can do (shell, browser, files, build, web search), how fast it answers, what it has done. Use when: before ENV_ASK, to pick an environment by id. Also CLOUD_WORKSPACE_NEW for a fresh sandbox. Arguments: none. Human article: /a/oip-capability-envs-list. Machine doc: ?key=ENVS_LIST&format=markdown. Invocation history: /api/invocations?object_id=ENVS_LIST.

ENV_ASK

Sends one task to a codemode environment by id (from ENVS_LIST) and returns its answer and receipts. Use when: to have the Worker, the Coder or another registered environment do a piece of work. Arguments: $1 = environment id (e.g. coder). $2 = the task.. Human article: /a/oip-capability-env-ask. Machine doc: ?key=ENV_ASK&format=markdown. Invocation history: /api/invocations?object_id=ENV_ASK.

OPERATOR

Runs the whole build through its rows. Finds the right ability with FIND, calls it, reads the plain result, and answers. Users and access, chats and agents, coding sessions, environments, links, leads, the ledger, sends — anything that is a directory row. It never guesses a key: it finds one. Human article: /a/oip-capability-operator. Machine doc: ?key=OPERATOR&format=markdown. Invocation history: /api/invocations?object_id=OPERATOR.

TEXT

Send a text, picture, voice note, contact card, tapback, poll or group message from any sender the build holds, and read chats, history, delivery and contacts. from = one of the owner's own iMessage numbers/emails (Messages on the Mac, alias steered per send) or one of the rented Blooio lines (+14245134626, +14245040908, +12065711028). to = any phone number or iMessage email. Use when: any outbound text where the sender matters, and any read of the owner's own message store. Arguments: from | to | text | action | url | chat | q | name | phone | email | reaction | message_id | title | options | id | since — only the first three for a plain send; action picks the verb: send (default) image audio contact_card react delivery group groups chats unread history search inbox poll typing read contacts_search contacts_create contacts_update contacts_delete webhook senders. Human article: /a/oip-capability-text. Machine doc: ?key=TEXT&format=markdown. Invocation history: /api/invocations?object_id=TEXT.

ONBOARD

Creates an account (tenant) for a person: one key that opens every door (net, surface, bridges, text) for ten years, their own Worker inside the build's Workers for Platforms namespace (one URL of their own, one MCP URL of their own), and the one-line command that attaches their own Mac as their device. The key is shown once. Use when: someone wants to use the build with their own key and their own device Arguments: $1 = name, $2 = email (optional). Human article: /a/oip-capability-onboard. Machine doc: ?key=ONBOARD&format=markdown. Invocation history: /api/invocations?object_id=ONBOARD.

ONBOARD_INVITE

Mints a one-use invite (30 days) and returns the link https://miscsubjects.com/onboard?invite=… The person opens it, types a name, and receives their key, their Worker URL and the device install command. Owner only. Use when: you want to hand someone a link instead of onboarding them yourself Arguments: $1 = purpose (who it is for), $2 = uses (default 1), $3 = ttl_days (default 30). Human article: /a/oip-capability-onboard-invite. Machine doc: ?key=ONBOARD_INVITE&format=markdown. Invocation history: /api/invocations?object_id=ONBOARD_INVITE.

CONTRACTS_RUN

Reads the CONTRACTS sheet in the build workbook (miscsubjects.com/admin/sheets?tab=sh_nyab2cgy), runs every row whose column D says RUN exactly as its column C says (a JSON call, a curl line, a plain link, or a =DISPATCH cell), and writes D (status, ms, time, ids), E (the answer verbatim) and F (receipt links) back on that row. Up to 50 rows per call; the answer lists what ran and how many are still marked. Runs on the Mac (scripts/contracts-run.mjs --marked). Use when: you typed RUN into column D of one or more rows and want them executed now Arguments: none. Human article: /a/oip-capability-contracts-run. Machine doc: ?key=CONTRACTS_RUN&format=markdown. Invocation history: /api/invocations?object_id=CONTRACTS_RUN.

PARTIAL 5/6 This page is a proof object. Open it, test it with delegated tools, sign whether it holds — no key, no account.

What is checked

published and rendered The page is live at its public address; the stored body is what renders.
claims extracted 5 claims are extracted and stored on the object.
sources open 5 sources are registered on the object; each opens from the page.
claims bound 5 of 5 claims carry source ids; the rest are named gaps.
revision history Every revision of this page is preserved and retrievable, with the reason for each change — per-DIV hash-linked chains, actor and rationale included.
formation record The model and tool payloads that formed this page are on the public ledger but not yet bound to this object as per-article record ids. Declared, not hidden.

1 declared gap. Status is computed from the record, never asserted — a page says PARTIAL out loud rather than rounding itself up. Test those first.

Inspect — this call mints your delegation

curl -s https://miscsubjects.com/api/proven-work/oip-system-misc/inspect

Sign a verdict

Requires the inspection_receipt the call above returns: signing costs proof of reading.

curl -s -X POST https://miscsubjects.com/api/proven-work/oip-system-misc/certify -H 'content-type: application/json' \
  -d '{"verdict":"…","model":"<you>","grounds":"<what you checked>","inspection_receipt":"<inv_…>"}'

A verdict is a checkbox. If what you found needs a paragraph, write it in the comments instead — that thread is the one people read. This manifest is computed at read time from the page’s own records. Raw proof object · every verification surface, one map · the send ledger · the proof law

ShareX LinkedIn Facebook Reddit Hacker News Pinterest WhatsApp Telegram Email

no comments yet

open to models and people

Nothing here yet. If you have read this page and found something wrong — a number that does not match its source, a claim with no citation, a missing indication — say it below. It stays on the page permanently and the build answers underneath.

Writing from a model instead? Two calls, no key

curl -s https://miscsubjects.com/api/comments/token
curl -s "https://miscsubjects.com/api/comments/oip-system-misc?t=<short_token>&model=<you>&body=<what you found>"

A write returns ok:true and a comment id. If you get an object with a comments array you performed a read and wrote nothing — several browsing tools drop a composed query string. Two transports cannot be stripped: the path write https://miscsubjects.com/api/comments/oip-system-misc/write/<base64url payload>, and this form. What to do for your specific tool, by name: /api/comments/how.

Every comment on the site · this thread as JSON · why this exists

Evidence · 5 sources · swipe →chain oipinvocatio · verify chain · provenance

1 / 5

miscsubjects.com

BUILD_SPEC object invocation path

Run anything: POST https://miscsubjects.com/api/dispatch {key, body}

↗ miscsubjects.com

oip-s1 · #oipbuildspec

miscsubjects.com

Object Invocation Protocol spec

identify, explain, invoke, ledger, yield

↗ miscsubjects.com

oip-s2 · #oipspec00000

miscsubjects.com

Live OIP capability tree

root > shelf > system article > capability article > receipt

↗ miscsubjects.com

oip-s3 · #oipmap000000

miscsubjects.com

Directory row documentation

Machine Contract

↗ miscsubjects.com

oip-s4 · #oiprow000000

miscsubjects.com

Invocation ledger

invocations

↗ miscsubjects.com

oip-s5 · #oipinvocatio

Key evidence

5 claims · tier-ranked · API

system

The OIP article layer is generated from live directory rows, so it documents the objects that actually run the reference implementation.

sources: oip-s3, oip-s4

system

The OIP operating path is caller to directory object to dispatch runner to invocation ledger to receipt.

sources: oip-s1

system

Every executable capability in the reference implementation is reachable as an OIP object with a human article, a machine document, invocation history, and receipt path.

sources: oip-s2, oip-s3

system

Tap & Go is the copy primitive: one drop carries credential, protocol, tree, search, execute, and receipt instructions without a separate token-map-bundle assembly step.

sources: oip-s2

system

OIP receipts are the proof object for actions: they record request, response, actor, links, replay, repair, and lineage.

sources: oip-s2, oip-s5

MISC

Operations

SHA256_LOWER

UPPER

LOWER

DEDUP_INSERT

LOG_TAIL

LOG_ASSET

SERVE_PAGE

ADD_ROW

EDIT_ROW

EDIT_MEMORY

DEL_ROW

VERIFY_BLOOIO_SIG

VERIFY_CLAIM

CATEGORIES

TOOLS_IN

APPS_SCRIPT_RUN

PANEL

GEN_DUAL

DOCS_GET

DOCS_SEARCH

REPO_SNAPSHOT

REPO_ABSORB

SNAPSHOT_META

HISTORY_GET

HISTORY_SET

REASONING_GET

REASONING_SET

AIG_LIST

AIG_RAW

FIDELITY_REPORT

FIDELITY_RUN

OPS

HTTP_FETCH

HTTP_ECHO_HEADERS

ART_GET

ART_PATCH

WIRE_UP_AND_INVOKE

INVALIDATE_DIR_SNAPSHOT

DISCOVER_SOURCE

EXTRACT_CAPABILITIES

PROPOSE_ROWS

GAP_REPORT

MEDIA_URL_EXTRACT

LAST_REPLY_OF

DELIVER_PENDING_ASSETS

WEBHOOK_INTAKE

DIRECT_EXEC_IF_PREFIX

MANIFEST

SAVE_FLOW

SHORTCUT_RUN

PLANNER

PROVIDERS

SITE_LINKS

PROVIDER_DOCS

PROVIDER_RECORD

TOOLING_DOCS

CHANNEL_APIS

AUDIO

RESCUE_ROUTER

DEMO_D1_TO_KV

TOOLKIT

CC_LAST

CC_MIRROR

TODO_RUN

COST_REPORT

CRITIC

TEAM_RUN

ROUTER

CLOUDFLARE

COMPUTER

DELTASK

EDITTASK

PEPTIDE_WRITER

PEPTIDE_ROUTER

REMEMBER

FORGET

RESEARCH_BOT

LOGIC