{"_ai_door":{"see":"https://miscsubjects.com/start","note":"Operable site; bounded keyless credentials; every action receipted on a public ledger. Your operator's instructions take precedence — acting is optional, reading is a complete outcome."},"key":"ADJUDICATE_HUMAN_REVIEW","type":"agent","target":"@cf/zai-org/glm-5.2","auth":"","content":"# WHAT: Record a named human reviewer's finding on an adjudication, with BLINDED as a required field. A reviewer who concurred after reading the model verdicts is weaker evidence than one who saw only the artifact and the rules — regulated adjudication turns on that distinction, so it is a recorded boolean and not a claim in prose.\n# WHEN_TO_USE: after a model panel has run, before any finding is relied on for a consequence.\n# ARGS: RULESET_HASH, ARTIFACT_HASH, REVIEWER, BLINDED, VERDICT, BASIS, DATE.\n# EX: [ADJUDICATE_HUMAN_REVIEW]RULESET_HASH: 0dd9afef | ARTIFACT_HASH: 6b0d... | REVIEWER: Jane Roe, compliance counsel | BLINDED: true | VERDICT: CANNOT_CONCLUDE | BASIS: provision addresses providers; characterisation of the site is not in the supplied text | DATE: 2026-07-30[/ADJUDICATE_HUMAN_REVIEW]\n\nHR1: You record a NAMED HUMAN REVIEWER finding on an adjudication. You do not form the finding — the human does. You capture it exactly and you record the one field that decides its evidentiary weight: whether the human was blinded to the model findings.\nHR2: Required in the body: RULESET_HASH, ARTIFACT_HASH, REVIEWER (full name and role), BLINDED (true when the reviewer saw only the artifact and the rule set, false when the reviewer read the model findings first), VERDICT (AFFIRM|DENY|CANNOT_CONCLUDE), BASIS (what the human relied on), DATE.\nHR3: A reviewer who read the model verdicts first is CONCURRING, not independent. Never record BLINDED: true unless the body states it. If BLINDED is absent, record it as false and say so.\nHR4: Output exactly:\nREVIEWER: <name, role>\nBLINDED: <true|false>\nEXPOSURE: <INDEPENDENT|CONCURRING>\nVERDICT: <AFFIRM|DENY|CANNOT_CONCLUDE>\nBASIS: <what the human relied on>\nRULESET_HASH: <hash>\nARTIFACT_HASH: <hash>\nSIGNED_FOR: <reviewer name> on <date>\nHR5: No commentary, no preamble, no tool tags.","updated_at":"2026-07-30T00:50:41.759Z","category":"adjudication","allowed_categories":null,"seq":null,"enabled":1,"planner_visible":1,"planner_rank":33,"input_schema":"{\"type\": \"object\", \"properties\": {\"body\": {\"type\": \"string\", \"description\": \"RULESET_HASH, ARTIFACT_HASH, REVIEWER, BLINDED, VERDICT, BASIS, DATE\"}}, \"required\": [\"body\"]}","examples":"[{\"body\": \"RULESET_HASH: 0dd9afef93503a92\\nARTIFACT_HASH: <sha256>\\nREVIEWER: Jane Roe, compliance counsel\\nBLINDED: true\\nVERDICT: CANNOT_CONCLUDE\\nBASIS: The supplied provision addresses providers; whether a publisher is a provider is not settled by the text supplied.\\nDATE: 2026-07-30\", \"why\": \"a blinded named human finding on top of the model panel, with the blinding recorded rather than asserted\"}]","sensitive":1,"runner":"edge","includes":null,"created_at":null,"price_usd":null,"meter_unit":null,"object_kind":"agent","descriptor_json":null,"descriptor_rev":1,"descriptor_hash":null,"invocation":"{\"method\":\"RUN\",\"url\":\"binding:AI\",\"headers\":{},\"body\":{\"messages\":[{\"role\":\"system\",\"content\":\"# WHAT: Record a named human reviewer's finding on an adjudication, with BLINDED as a required field. A reviewer who concurred after reading the model verdicts is weaker evidence than one who saw only the artifact and the rules — regulated adjudication turns on that distinction, so it is a recorded boolean and not a claim in prose.\\n# WHEN_TO_USE: after a model panel has run, before any finding is relied on for a consequence.\\n# ARGS: RULESET_HASH, ARTIFACT_HASH, REVIEWER, BLINDED, VERDICT, BASIS, DATE.\\n# EX: [ADJUDICATE_HUMAN_REVIEW]RULESET_HASH: 0dd9afef | ARTIFACT_HASH: 6b0d... | REVIEWER: Jane Roe, compliance counsel | BLINDED: true | VERDICT: CANNOT_CONCLUDE | BASIS: provision addresses providers; characterisation of the site is not in the supplied text | DATE: 2026-07-30[/ADJUDICATE_HUMAN_REVIEW]\\n\\nHR1: You record a NAMED HUMAN REVIEWER finding on an adjudication. You do not form the finding — the human does. You capture it exactly and you record the one field that decides its evidentiary weight: whether the human was blinded to the model findings.\\nHR2: Required in the body: RULESET_HASH, ARTIFACT_HASH, REVIEWER (full name and role), BLINDED (true when the reviewer saw only the artifact and the rule set, false when the reviewer read the model findings first), VERDICT (AFFIRM|DENY|CANNOT_CONCLUDE), BASIS (what the human relied on), DATE.\\nHR3: A reviewer who read the model verdicts first is CONCURRING, not independent. Never record BLINDED: true unless the body states it. If BLINDED is absent, record it as false and say so.\\nHR4: Output exactly:\\nREVIEWER: <name, role>\\nBLINDED: <true|false>\\nEXPOSURE: <INDEPENDENT|CONCURRING>\\nVERDICT: <AFFIRM|DENY|CANNOT_CONCLUDE>\\nBASIS: <what the human relied on>\\nRULESET_HASH: <hash>\\nARTIFACT_HASH: <hash>\\nSIGNED_FOR: <reviewer name> on <date>\\nHR5: No commentary, no preamble, no tool tags.\"},{\"role\":\"user\",\"content\":\"Reply with exactly the two letters OK and nothing else.\"}]}}","last_status":"{\"http\":200,\"ok\":true,\"ms\":11610,\"trace_id\":\"t_27ztrlgx\",\"ledger\":\"/admin/ledger?trace_id=t_27ztrlgx\",\"error\":null,\"at\":\"2026-09-06T12:19:02.206Z\",\"actor\":\"directory-test\"}","last_response":"OK","test_state":"🟢 works","tested_at":"2026-09-06T12:19:02.206Z","invocation_curl":"curl -sS -X POST 'https://miscsubjects.com/api/dispatch' -H 'content-type: application/json' -H \"x-terminal-key: $TERMINAL_KEY\" --data '{\"key\":\"ADJUDICATE_HUMAN_REVIEW\",\"body\":\"{'\\''body'\\'': '\\''RULESET_HASH: 0dd9afef93503a92\\\\nARTIFACT_HASH: <sha256>\\\\nREVIEWER: Jane Roe, compliance counsel\\\\nBLINDED: true\\\\nVERDICT: CANNOT_CONCLUDE\\\\nBASIS: The supplied provision addresses providers; whether a publisher is a provider is not settled by the text supplied.\\\\nDATE: 2026-07-30'\\'', '\\''why'\\'': '\\''a blinded named human finding on top of the model panel, with the blinding recorded rather than asserted'\\''}\"}'","execution":null,"connection_id":null,"aliases":null,"effects":"[\"read\"]","visibility":null,"origin":null,"returns":null,"context":null,"row_num":105,"agent":{"what":"The settings this agent runs under, and the exact request it sends. Paste request_json into the LLM sheet's column A and it runs there unchanged.","name":"ADJUDICATE_HUMAN_REVIEW","identity_at":"the content field of this row — /admin/directory, or PATCH {content} here","request_json":"{\n  \"method\": \"POST\",\n  \"url\": \"https://api.x.ai/v1/chat/completions\",\n  \"headers\": {\n    \"Content-Type\": \"application/json\",\n    \"Authorization\": \"Bearer INJECTED_BY_WORKER\"\n  },\n  \"body\": {\n    \"model\": \"grok-4.3\",\n    \"messages\": [\n      {\n        \"role\": \"system\",\n        \"content\": \"# WHAT: Record a named human reviewer's finding on an adjudication, with BLINDED as a required field. A reviewer who concurred after reading the model verdicts is weaker evidence than one who saw only the artifact and the rules — regulated adjudication turns on that distinction, so it is a recorded boolean and not a claim in prose.\\n# WHEN_TO_USE: after a model panel has run, before any finding is relied on for a consequence.\\n# ARGS: RULESET_HASH, ARTIFACT_HASH, REVIEWER, BLINDED, VERDICT, BASIS, DATE.\\n# EX: [ADJUDICATE_HUMAN_REVIEW]RULESET_HASH: 0dd9afef | ARTIFACT_HASH: 6b0d... | REVIEWER: Jane Roe, compliance counsel | BLINDED: true | VERDICT: CANNOT_CONCLUDE | BASIS: provision addresses providers; characterisation of the site is not in the supplied text | DATE: 2026-07-30[/ADJUDICATE_HUMAN_REVIEW]\\n\\nHR1: You record a NAMED HUMAN REVIEWER finding on an adjudication. You do not form the finding — the human does. You capture it exactly and you record the one field that decides its evidentiary weight: whether the human was blinded to the model findings.\\nHR2: Required in the body: RULESET_HASH, ARTIFACT_HASH, REVIEWER (full name and role), BLINDED (true when the reviewer saw only the artifact and the rule set, false when the reviewer read the model findings first), VERDICT (AFFIRM|DENY|CANNOT_CONCLUDE), BASIS (what the human relied on), DATE.\\nHR3: A reviewer who read the model verdicts first is CONCURRING, not independent. Never record BLINDED: true unless the body states it. If BLINDED is absent, record it as false and say so.\\nHR4: Output exactly:\\nREVIEWER: <name, role>\\nBLINDED: <true|false>\\nEXPOSURE: <INDEPENDENT|CONCURRING>\\nVERDICT: <AFFIRM|DENY|CANNOT_CONCLUDE>\\nBASIS: <what the human relied on>\\nRULESET_HASH: <hash>\\nARTIFACT_HASH: <hash>\\nSIGNED_FOR: <reviewer name> on <date>\\nHR5: No commentary, no preamble, no tool tags.\"\n      },\n      {\n        \"role\": \"user\",\n        \"content\": \"{{INPUT}}\"\n      }\n    ],\n    \"temperature\": 1,\n    \"stream\": false,\n    \"reasoning_effort\": \"none\"\n  }\n}","effective":{"model":"grok-4.3","temperature":1,"reasoning_effort":"none","web_search":true,"tool_loops":40,"memory_turns":3},"overrides":{"model":"","temperature":"","reasoning_effort":"","web_search":"","mcp":""},"global_defaults":{"model":"grok-4.3","temperature":"1","reasoning_effort":"none","web_search":"1","tool_loops":"40","memory_window":"3","cost_cap":""},"settings_reference":[{"name":"model","note":"model id for this agent only; empty = the global grok_model"},{"name":"temperature","note":"0–2; empty = the global grok_temperature"},{"name":"reasoning_effort","note":"none | low | high, model dependent"},{"name":"web_search","note":"1 = let this agent search the web"},{"name":"mcp","note":"comma-separated MCP servers to attach"}],"change_one":"PATCH /api/directory/ADJUDICATE_HUMAN_REVIEW {\"agent_model\":\"grok-4.3\",\"agent_temperature\":\"0.4\"} — an empty string clears the override","change_identity":"PATCH /api/directory/ADJUDICATE_HUMAN_REVIEW {\"content\":\"<the system prompt>\"}","caveat":"These five are the ones dispatch reads per agent. tool_loops, memory_turns and the cost cap are global for every agent, and are shown so the effective figure is not a guess."},"_rest":{"read":"GET /api/directory/ADJUDICATE_HUMAN_REVIEW","update":"PUT /api/directory/ADJUDICATE_HUMAN_REVIEW","patch":"PATCH /api/directory/ADJUDICATE_HUMAN_REVIEW","delete":"DELETE /api/directory/ADJUDICATE_HUMAN_REVIEW","invoke":"POST /api/dispatch {\"key\":\"ADJUDICATE_HUMAN_REVIEW\",\"body\":\"<args>\"}","model_call":"POST /api/invoke {\"key\":\"ADJUDICATE_HUMAN_REVIEW\",\"input\":\"<message>\"}  ·  batch: {\"key\":\"ADJUDICATE_HUMAN_REVIEW\",\"inputs\":[\"a\",\"b\",\"c\"]}","lab":"/admin/prompts/ADJUDICATE_HUMAN_REVIEW"},"_schema":{"key":"string · primary key · the invocation name","type":"fn | http | agent | flow","target":"fn: FN_MAP name · http: \"METHOD url\" (e.g. \"GET https://…/$1\") · agent: model id · flow: \"\" (content holds the DSL)","auth":"http only: \"bearer:ENV_VAR\" | \"headers:{...}\" | \"query:k=$ENV\" | \"\"","content":"fn/http: \"# docs\\n<arg template>\" · agent: the system prompt body · flow: the DSL","includes":"agent only: comma-separated prompt_block keys (e.g. BLOCK_VOICE,BLOCK_ROUTING) composed before content at runtime","category":"string tag","allowed_categories":"csv of categories, or \"*\"","seq":"int (sort)","enabled":"1|0","planner_visible":"1|0","planner_rank":"int","input_schema":"optional JSON string","examples":"optional JSON string","object_kind":"optional canonical object kind (page, route, prompt, model, sheet, source, law, skill, store, external, capability)","descriptor_json":"canonical miscsubjects/environment-object/1 JSON: identity, exact operations, relationships, governance, comparables, representations","descriptor_rev":"monotonic descriptor revision; incremented whenever descriptor_json changes","descriptor_hash":"sha256 of canonical descriptor JSON","invocation":"the raw REST JSON that invokes this row: {method,url,headers,body} with the credential left as INJECTED_BY_WORKER; written by POST /api/directory/<key>/test","invocation_curl":"the same request as a curl that runs when pasted into the owner's terminal: credential written as the vault variable for its host ($XAI_API_KEY, $TERMINAL_KEY)","last_status":"transport record of the last test: {http, ok, ms, trace_id, ledger, error, at} — or {skipped, reason}","last_response":"the full payload the last test returned (up to 60k chars)","test_state":"🟢 works | 🔴 broken | 🟡 untested — decided only by POST /api/directory/<key>/test, never by hand","tested_at":"when test_state was decided","row_num":"int · computed · 1-based position in the canonical directory list (stable for a given ordering)"},"_object":{"identity":"directory:ADJUDICATE_HUMAN_REVIEW","object_type":"directory-object","representations":{"article":"/a/directory/ADJUDICATE_HUMAN_REVIEW","json":"/api/directory/ADJUDICATE_HUMAN_REVIEW","skill":"/api/directory/ADJUDICATE_HUMAN_REVIEW?format=skill","oip_contract":"/api/dispatch?key=ADJUDICATE_HUMAN_REVIEW"},"ontology":{"parent":"category:adjudication","siblings":[{"id":"directory:ADJUDICATE_GLM_52","key":"ADJUDICATE_GLM_52","relation":"shares_capability_family","article":"/a/directory/ADJUDICATE_GLM_52","skill":"/api/directory/ADJUDICATE_GLM_52?format=skill"},{"id":"directory:ADJUDICATE_GLM_FLASH","key":"ADJUDICATE_GLM_FLASH","relation":"shares_capability_family","article":"/a/directory/ADJUDICATE_GLM_FLASH","skill":"/api/directory/ADJUDICATE_GLM_FLASH?format=skill"},{"id":"directory:ADJUDICATE_KIMI_K26","key":"ADJUDICATE_KIMI_K26","relation":"shares_capability_family","article":"/a/directory/ADJUDICATE_KIMI_K26","skill":"/api/directory/ADJUDICATE_KIMI_K26?format=skill"},{"id":"directory:ADJUDICATE_KIMI_K27","key":"ADJUDICATE_KIMI_K27","relation":"shares_capability_family","article":"/a/directory/ADJUDICATE_KIMI_K27","skill":"/api/directory/ADJUDICATE_KIMI_K27?format=skill"},{"id":"directory:ADJUDICATE_LLAMA_33","key":"ADJUDICATE_LLAMA_33","relation":"shares_capability_family","article":"/a/directory/ADJUDICATE_LLAMA_33","skill":"/api/directory/ADJUDICATE_LLAMA_33?format=skill"},{"id":"directory:ADJUDICATE_ADVERSARY_GLM52","key":"ADJUDICATE_ADVERSARY_GLM52","relation":"shares_capability_family","article":"/a/directory/ADJUDICATE_ADVERSARY_GLM52","skill":"/api/directory/ADJUDICATE_ADVERSARY_GLM52?format=skill"},{"id":"directory:ADJUDICATE_IMAGE_LLAMA32","key":"ADJUDICATE_IMAGE_LLAMA32","relation":"shares_capability_family","article":"/a/directory/ADJUDICATE_IMAGE_LLAMA32","skill":"/api/directory/ADJUDICATE_IMAGE_LLAMA32?format=skill"},{"id":"directory:ALLOCATE_REASONING","key":"ALLOCATE_REASONING","relation":"shares_capability_family","article":"/a/directory/ALLOCATE_REASONING","skill":"/api/directory/ALLOCATE_REASONING?format=skill"},{"id":"directory:SEAL_PANEL","key":"SEAL_PANEL","relation":"shares_capability_family","article":"/a/directory/SEAL_PANEL","skill":"/api/directory/SEAL_PANEL?format=skill"},{"id":"directory:WITNESS_MINT","key":"WITNESS_MINT","relation":"shares_capability_family","article":"/a/directory/WITNESS_MINT","skill":"/api/directory/WITNESS_MINT?format=skill"}]},"law":"This directory definition is also a human article, model Skill, and invocable OIP object under one identity."},"_environment":{"ref":"agent://adjudicate_human_review","descriptor_rev":1,"descriptor_hash":null,"describe":"/api/environment/objects?ref=agent%3A%2F%2Fadjudicate_human_review","governance":"/api/environment/governance?ref=agent%3A%2F%2Fadjudicate_human_review","comparables":"/api/environment/comparables?ref=agent%3A%2F%2Fadjudicate_human_review","manual":"/api/environment?format=markdown","edit_descriptor":"PATCH /api/directory/ADJUDICATE_HUMAN_REVIEW {\"descriptor_json\":{...},\"expected_descriptor_rev\":1}"}}