← ClaudeAtlas

memex-updatelisted

Refresh an existing Memex from deterministic Git, wiki, or authorized-source changes while preserving unaffected content. Use when a user asks to update stale documentation, reconcile changed code, run a no-op freshness check, or finalize new source evidence.
giulioleone097/memex · ★ 0 · AI & Automation · score 70
Install: claude install-skill giulioleone097/memex
# Update Memex Detect changes first. Write only affected pages and record a run only after successful validation. ## Host paths - Codex: derive the plugin root from this SKILL.md path by removing `/skills/memex-update/SKILL.md`; never derive it from the current working directory. - Claude Code: use `${CLAUDE_PLUGIN_ROOT}` as the plugin root. - Define `<cli>` as `node "<plugin-root>/dist/cli.js"` for Codex or `node "${CLAUDE_PLUGIN_ROOT}/dist/cli.js"` for Claude Code. ## Preconditions - Require a healthy initialized wiki and explicit mode/root. - Record current `status`, content hash, last finalized Git HEAD, and dirty working-tree state before writes. - For external-source changes, require already-ingested, redacted evidence; never fetch a connector implicitly. ## Procedure 1. Run `status` using the same mode/root. In code mode, delegate graph freshness and changed-path mapping to `memex-graph`'s `changes` action; that skill owns graph status, authorized refresh, limits, and confidence reporting. After an authorized graph build (and, when the concept/wiki plane exists, `enrich`), run `memex-graph`'s `report` action so `graph-report.md` reflects the current unified graph; treat its output as one of the changed pages for this run. In personal mode, skip graph and report work. 2. Run `context` using the same mode/root. 3. If context reports no changed evidence, run `check`, return a no-op result, and do not call `write`, `enrich`, or `finalize`. 4. Map changed evidence to