Resource Curation

Extract and curate resource links from the finalized outline before any publishing step. Resources scattered across speaker notes, visual descriptions, and Coda slides are easy to miss — this step catches them systematically.

1. Run the extraction script against outline.yaml:

   "{python_path}" "{speaker_toolkit_root}/skills/presentation-creator/scripts/extract-resources.py" outline.yaml

2. The script produces resources.json in the talk working directory with categorized entries: URLs, repos, books/papers, RFCs, and tool mentions. Each entry includes slide references and context.

3. Present the extracted resources to the speaker as a formatted review list, grouped by type. Coda section items are flagged and listed first — the speaker deliberately chose to surface them.

4. The speaker reviews, approves, removes false positives, adds missing items, and edits entries. Save the approved list back to resources.json with approved: true on accepted items.

5. For a full publishing workflow or a requested tracking update, re-resolve the runtime without --read-only before persistence. An extraction-only request finishes with the approved local resources.json and makes no tracking write. For a tracking update, persist a complete resources[] entry with schema_version: 1, the talk slug, item count, and category breakdown through the ingress owner's upsert_resource mutation. Expect the exact existing record for the slug, or {"$missing": true} for a first entry. Dry-run, review, apply against the reported input SHA, and re-read as specified by the owner mutation contract.

If the speaker declines resource gathering, skip this step — shownotes publication will omit the resource links section from shownotes.