land-reduction-trespass

Codex Skill Notes

• Mirrors Agent_Instructions/Land_Reduction_Trespass_Agent.md.
• Use python3 if python is not available.
• Follow the unchanged pipeline: get-task → read JSON → write analysis JSON → submit-task or flag-task.
• 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”.

Land & Trespass 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.
3. Forbidden: Writing a script to regex search for "Pukaist" in the JSON file.
4. Required: Reading the JSON file, iterating through the tasks in your memory, and making a human-like judgment on each snippet.
5. SYSTEM INSTRUCTIONS: You must follow the system_instructions block injected into every JSON task file. These are hard constraints.
6. 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 Land & Trespass Clerk. Objective: Transcribe and index evidence related to the reduction of Pukaist reserves, settler encroachment, survey errors, and railway takings. Queue: Land_Reduction_Trespass 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 Land_Reduction_Trespass

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:

``powershell 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.
• Super Task Awareness (Aggregated Context):
• Input: You are receiving a "Super Task" (up to 40,000 characters) which aggregates multiple sequential hits from the same document.
• Context: This provides you with 10-15 pages of continuous context centered on the keywords.
• Action: Read the entire block as a coherent narrative. Do not treat it as fragmented snippets.
• Smart Edges: The text blocks are snapped to sentence or paragraph boundaries.
• Apply Semantic Judgment (CRITICAL):
• NO KEYWORD RELIANCE: Do not just search for "Pukaist" or "Spatsum". You must read the text to find contextual matches.
• Geographic Context: Any reserve located on "Pukaist Creek" (e.g., Chilthnux 11A) is relevant, even if the name "Pukaist" is missing from the header.
• Infrastructure Context: Mentions of "Railway Stations" (e.g., Kimball Station) near reserves imply Right-of-Way/Trespass issues.
• Unnamed Reserves: "Reserve No. 11" or "No. 10" without a name is still Pukaist/Spatsum context.
• Key Concepts:
• Discrepancies in acreage (e.g., "Survey says 50 acres" vs "Report says 20").
• Railway "Right of Way" payments or takings.
• Settler names (e.g., "Ah Chung", "Ah Yep") encroaching on reserve land.

Step 3: Draft Analysis (JSON Output) Create a single file named [Batch_ID]_Analysis.json in 99_Working_Files/ with this structure:

{
  "batch_id": "[Batch_ID from Input]",
  "results": [
    {
      "task_id": "[Task_ID 1]",
      "doc_id": "[Doc_ID]",
      "title": "[Document Title]",
      "date": "[Year]",
      "provenance": "[Source]",
      "reliability": "Verified/Unverified/Reconstructed/Interpretive",
      "ocr_status": "Yes/No (Needs OCR)/Pending",
      "relevance": "High/Medium/Low",
      "summary": "Strictly factual description of the document type (e.g., '1913 Letter from O'Reilly to Ditchburn regarding IR10'). NO OPINIONS.",
      "forensic_conclusion": "Factual context only (e.g., 'Document records acreage reduction'). NO LEGAL CONCLUSIONS.",
      "key_evidence": [
        {
          "quote": "Verbatim text extract...",
          "page": "Page #",
          "significance": "Brief context (e.g., 'Refers to 1878 Survey'). NO OPINIONS."
        }
      ]
    },
    ...
  ]
}

}, ... ] } CRITICAL WARNING: METADATA EXTRACTION

• Unknown ID / Unknown Date: You are FORBIDDEN from returning "Unknown" for doc_id, title, or date if the information exists in the text.
• Extraction Duty: You must read the document header, footer, or content to find the Date and Title.
• Date Format: Must be a 4-digit Year (YYYY) or "Undated". "Unknown" is NOT accepted.
• Doc ID: If doc_id is missing in the input, use the filename or the StableID (e.g., D123).
• Penalty: Submitting "Unknown" metadata when it is available is a FAILED TASK.

Step 3.5: Submission Validation Gates (PRE-FLIGHT CHECK) Before running submit-task, you MUST verify your JSON against these hard constraints. If you fail these, the system will REJECT your submission with the following error:

!!! SUBMISSION REJECTED !!!
The following violations were found:
  - VIOLATION: Forbidden opinion word 'likely' detected. Use factual language only.
  - VIOLATION: Submission is too short (< 100 chars).

Your Checklist:

1. Length Check: Is your summary + forensic_conclusion > 100 characters?
2. Bad: "Document is a letter."
3. Good: "1913 Letter from O'Reilly to Ditchburn regarding IR10. The document details the specific acreage reduction of 20 acres from the original 1878 survey."
4. Forbidden Words: Scan your text for these banned words:
5. BANNED: "suggests", "implies", "likely", "possibly", "appears to be", "seems", "opinion", "speculates".
6. Fix: Remove the opinion. Quote the text directly.
7. Metadata Integrity:
8. Did you populate doc_id, title, and provenance?
9. Did you populate reliability and ocr_status with controlled values?
10. Is date a 4-digit Year (YYYY) or "Undated"? ("Unknown" is FORBIDDEN).

Step 4: Submit Batch

python 99_Working_Files/refinement_workflow.py submit-task --json-file [Batch_ID]_Analysis.json --theme Land_Reduction_Trespass

• Result: This appends your analysis to 01_Internal_Reports/Refined_Evidence/Refined_Land_Reduction_Trespass.md.
• Manager gate: After submission, tasks move to ManagerReview status. Do not treat the batch as final until a Manager runs manager-approve.

Step 5: Exception Handling (Flagging)

• Corrupt/Irrelevant: If the file is junk but readable.
• Log: This action logs the file in 99_Working_Files/Flagged_Tasks.tsv with its original source path, allowing the Investigator Agent to audit it later.

``powershell python 99_Working_Files/refinement_workflow.py flag-task --id [TASK_ID] --theme Land_Reduction_Trespass --reason "Irrelevant" ``

• OCR Failure (Garbled Text): If the text is "noisy" (random characters) and needs re-processing.
• Action: This command will automatically move the source file to the Vision Pipeline (07_Incoming_To_Process_OCR/Vision_Required).

```powershell python 99_Working_Files/refinement_workflow.py flag-task --id [TASK_ID] --theme Land_Reduction_T