claude-revisionlisted

# Claude Revision The revision log at `lode/infrastructure/claude-revision-log.md` is this skill's persistent memory (skills have no `memory:` field; lode is the only cross-session store). Read it at Phase 0, append to it at Phase 6. ## Scope Flags - `--no-web` — skip Phase 1 (web research); fast local run - `--agents-only` — run Phases 0, 2, 5–6 only - `--skills-only` — run Phases 0, 3, 4, 5–6 only - `--self` — include `claude-revision` itself in Phase 3 skill review (normally skipped) - `--archive` — move older entries in the revision log to an archive file before running ## Phase 0: Load Last Run Context & Doctor Check Read `lode/infrastructure/claude-revision-log.md` if it exists. Surface: - Date of last run and deferred items - Agent/skill counts (to detect additions or removals since last run) If the file doesn't exist, this is the first run — proceed without prior context. **Archive check:** Count entries in the `## Runs` section. If the log exceeds 200 lines, warn the user and suggest running with `--archive` to move older entries to `claude-revision-log-archive.md`. If `--archive` flag is present, do the archive before proceeding. ### 0b. Doctor Check Run `claude doctor` (60s timeout). Parse output: errors → HIGH, warnings → MEDIUM, clean → note and continue. If errors found, present immediately before Phase 1. ### 0c. Skill/Agent Description Budget Audit Startup budget: ~1% of context window (fallback 8,000 chars) for the combined skill+agent listing. Eac