knowledge-curatorlisted

# Skill: Knowledge Curator <!-- Knowledge source: GENERALIZE of the private EI's knowledge-curator SKILL.md. Thin pointer to playbooks/knowledge-curator.md - one canonical source per D-007. --> Full workflow: [`playbooks/knowledge-curator.md`](../../playbooks/knowledge-curator.md). ## When this applies Before a promotion batch, before packet planning as a freshness gate, after a packet closeout that changed methodology or skills, or when an agent keeps asking something the knowledge base should already answer. Not after every session. That is [`run-retro`](../run-retro/SKILL.md)'s cadence, and a different question. ## Process 1. Read the previous findings ledger first, so each signal can be classified as new, recurring or regressed rather than reported fresh every time. 2. Collect only signals that already exist: link check, knowledge index orphans, `review_after` freshness, frontmatter validation, privacy scan and its suppression hygiene, deferred closeout items, integration health. Generating new evidence means this stopped being a curator run. 3. Classify each finding: severity, one of the six canonical routing tokens, fix class A/B/C, and an evidence label. 4. Escalate anything open at three or more occurrences by one severity level, with a prevention proposal rather than another fix. 5. Write the report from [`templates/curator-report.md`](../../templates/curator-report.md). 6. Only with an explicit opt-in, apply Class A fixes on their own