What problem does it solve? AI-extracted knowledge entries are provisional until a human verifies them, and low-confidence or stale entries can pollute a codebase knowledge base. This Skill provides a structured review workflow to confirm, reject, or edit knowledge entries so the knowledge base stays accurate. ## Core Features & Use Cases - Entry Review Workflow: Confirm accurate entries (confidence raised to 1.0), reject inaccurate ones, or edit content in place via codekb review CLI commands or the codekb_review MCP tool. - Lifecycle Management: Track entry statuses including accepted, draft, potentially_stale, superseded, and rejected, with filtering via codekb list flags. - Git-Tracked Audit Trail: Every modification is captured in Git history under codekb/knowledge/, enabling full traceability of review decisions. - Use Case: After codekb extract reports low-confidence entries, list entries with potentially_stale status, verify each against the actual source code, then confirm or reject them before they are used in an OpenSpec proposal workflow. ## Quick Start Ask the AI to list all potentially stale CodeKB knowledge entries and review each one against the source code, confirming accurate entries and rejecting inaccurate ones.
