← ClaudeAtlas

memory-tag-lintlisted

The write gate for /encode-codebase — the seven assertions every candidate memory must pass before it is written, and the positive control that proves the gate is wired up. Auto-loads before encoding a unit, before re-checking an encoded batch, and when running the proving gate on a draft vocabulary. Load it whenever a tag is about to be written or judged.
konstantinos-malavazos/claude-code-playbook · ★ 2 · AI & Automation · score 68
Install: claude install-skill konstantinos-malavazos/claude-code-playbook
# Memory tag lint (the write gate) **This file is the single definition of the gate.** Three agents apply it and they must apply it identically: `@encoder` enforces it before every write, `@encode-rechecker` re-runs it as a positive-controlled audit, and `@encode-recon` checks its 10 dry-run samples the same way. If one of those files and this skill disagree, **this skill wins** — say so and stop, rather than enforcing two gates. That rule exists because of a real failure. A vocabulary cache key was renamed, the drifted-tag assertion silently dereferenced a key that no longer existed, and the check never fired again — for 29 tags across two vocabulary versions. Nothing errored. The gate reported clean the whole time. **A schema that no single file owns is a schema that drifts, and this one failed open.** ## Check against the cache, never against memory of it Every assertion is evaluated against the regenerated vocabulary cache at `<workspace>/.claude/encode-vocab/<repo>.json`, read at the time of the check. Never against what the vocabulary said earlier in the session, and never against a vocabulary you remember approving. The cache is a build artifact regenerated every run, and it is the only copy the gate reads. **A hand-edited cache is reported, not repaired.** Use it as it stands and put one line in the report. Repairing it locally hides a drift between the cache and its authoritative source. ## The seven assertions All seven pass → write. Any fail → fix the candid