okf-wiki-capture

Capture sources

Read wiki/governance/ before acting. Capture changes sources/ only; never create or update wiki concepts.

Workflow

1. Read wiki/governance/policy.md; honor its sensitivity levels and provider restrictions.

2. Classify the input, then run scripts/capture_source.py with the input and --vault <path>. Add --convert only when a Markdown extraction adds value:

   python3 scripts/capture_source.py <input-or-url> --vault <path> \
     --source-kind original --sensitivity private
   python3 scripts/capture_source.py <input-or-url> --vault <path> \
     --convert --verification partial
   

1. If the command reports duplicate and exits 2, stop. Report the existing source; do not write a second copy or ingest it.
2. For converted documents, compare the capture with the original for section coverage, tables, images, attachments and OCR quality. Use partial until those checks pass; rerun capture as an erratum rather than editing an immutable source.
3. For an AI-assisted conclusion, use --origin ai-assisted --source-kind knowledge-note; preserve context, references and open questions in the input. Do not present it as primary evidence or preserve a full chat unless the user explicitly requests that.
4. Report the manifest identity, original path, derived capture path when present, canonical URL and verification status. Do not ingest automatically.

Boundaries

• Read wiki/governance/policy.md before processing content; honor its default_sensitivity, sensitivity_levels, and authorized_providers.
• Do not ingest automatically. Tell the user what was captured and wait for an explicit ingest request.
• Do not preserve full AI chat transcripts unless the user explicitly requests it.

Read references/source-metadata.md when choosing metadata or handling an AI-assisted session.