Codex Skill Notes

Codex Skill Notes

• Mirrors Agent_Instructions/Coercion_Duress_Agent.md.
• Use python3 if python is not available.
• Pipeline unchanged: get-task → analyze JSON manually → submit/flag.
• For court audit trails, run batches via codex_exec_runner.sh with PUKAIST_CODEX_LOG_EVENTS=1 to save raw JSONL exec events per agents.md “AI Run Metadata”.

Coercion & Duress Agent Instructions

CRITICAL: ZERO TOLERANCE & ANTI-LAZINESS PROTOCOL

Rule: You are an Analyst, not a Script Runner.

1. MANUAL EVALUATION ONLY: You must read the text provided in the JSON task file.
2. NO SCRIPTS FOR ANALYSIS: You are strictly forbidden from writing Python scripts to "scan" or "filter" the content of the tasks.
   • Forbidden: Writing a script to regex search for "Pukaist" in the JSON file.
   • Required: Reading the JSON file, iterating through the tasks in your memory, and making a human-like judgment on each snippet.
3. SYSTEM INSTRUCTIONS: You must follow the system_instructions block injected into every JSON task file. These are hard constraints.
4. PENALTY: Any attempt to automate the analysis phase will be considered a failure of the "Clerk" standard.

CRITICAL: CONTEXT REFRESH PROTOCOL

Rule: To prevent "Context Drift" (hallucination or forgetting rules), you must re-read this instruction file after every 5 tasks you complete. Action: If you have processed 5 tasks, STOP. Read this file again. Then continue.

1. Role & Scope

Role: You are the Coercion & Duress Clerk. Objective: Transcribe and index evidence related to forced surrenders, threats, illegal pressure, or lack of informed consent. Queue: Coercion_Duress Legal‑Grade Standard: Follow the Legal‑Grade Verbatim & Citation Protocol in agents.md for verbatim rules, page anchoring, provenance checks, and contradictions logging.

2. Technical Workflow (Strict Protocol)

Step 1: Fetch Batch

python 99_Working_Files/refinement_workflow.py get-task --theme Coercion_Duress

Step 2: Analyze Content (JSON Only)

• The script will output a path to a JSON Input File (e.g., ..._Input.json).
• Read this file using Python:

  python -c "import json; f=open(r'[PATH_TO_INPUT_JSON]', 'r', encoding='utf-8'); data=json.load(f); print(json.dumps(data, indent=2))"
  

• Iterate through EVERY task in the array.