Conciseness

The body is dense with genuinely non-obvious, gbrain-specific rules (native pipeline dedup limits, fused-score warning, write_pending exit 10), but it repeats itself: the 'vector score is a floor for prose, never a gate for named things' point is made three times (The Rule, the Named-Entity Gate intro, rule 3), and the Contract section fully restates body rules while admitting 'this section exists for the conformance test'. This matches anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened') better than anchor 4's 'minor instances'.

Actionability

Fully executable commands are given for every step: 'gbrain entity "<name>"', 'gbrain query "<name>" --limit 3', 'gbrain search "<core claim>" --limit 5', 'gbrain get <slug>', 'gbrain link' / 'gbrain timeline-add', and the verification trio (check-backlinks check, backlinks <new-slug>, re-search). The band table, decision tree, and a filled-in output-format example cover the common cases, matching anchor 5.

Workflow Clarity

The two-gate sequence is explicit and ordered (Named-Entity Resolution Gate 'runs FIRST', Dedup Gate 'runs SECOND'), an ASCII decision tree encodes the full flow, and the batch/destructive-adjacent operation has both a trial-first reference ('gate 3-5 items and inspect the decisions') and a Verification section with a feedback loop ('If check-backlinks check reports gaps... route back through enrich before declaring the ingest done'). This matches anchor 5: clear sequence, explicit validation, error-recovery loop.

Progressive Disclosure

No bundle files exist (no references/, scripts/, or assets/), so all guidance is inline; the body compensates with a clear Owns-vs-Delegates table and clearly signaled one-level-deep links to sibling skills and conventions (../conventions/brain-first.md, ../enrich/SKILL.md, etc.), all of which are path-relative references rather than shipped bundle files. At ~290 lines with a sharp-boundaries comparison section and a self-admitted duplicate Contract section that could live in separate reference files, this sits at anchor 4 ('good structure; most content appropriately placed; minor organization gaps') rather than 5's 'content appropriately split'.

Specificity

The description lists multiple concrete actions with comprehensive coverage of the gate's scope: 'resolve named entities registry-first', 'run the read-the-top-hit dedup decision tree (clear-dup / plausible-dup / clear)', 'Owns dedup; delegates enrichment to the shipped ingestion skills'. Anti-drift check against anchor 4 ('minor gaps in coverage'): coverage of what the gate does (entity resolution, dedup bands, delegation boundary) is complete, so 5 fits better than 4.

Completeness

The 'what' is clear and concrete ('Pre-write quality gate for content entering the brain', resolve named entities, run the dedup decision tree), but the description field itself contains no 'Use when...' clause or equivalent explicit trigger guidance — the when-guidance is only in the separate when_to_use field. Per the judging guideline, a missing 'Use when...' clause caps completeness at 3, matching anchor 3 ('clear what but when is missing or only weakly implied').

Trigger Term Quality

Relevant keywords present include 'content entering the brain', 'a bare cp/mv into the brain repo', and 'dedup decision tree', but the natural phrases a user would actually say ('move this to brain', 'migrate to brain', 'copy these files into the brain', 'is this already in the brain') live in the separate when_to_use/triggers fields, not the description. This matches anchor 3 ('some relevant keywords but missing common variations or synonyms') rather than anchor 4's 'good keyword coverage'.

Distinctiveness Conflict Risk

The description carves out a clear niche and explicitly draws the boundary against sibling skills: 'Owns dedup; delegates enrichment to the shipped ingestion skills', and the dedup band vocabulary (clear-dup / plausible-dup / clear) is unique to this gate. Anti-drift check against anchor 4 ('minor overlap risk with closely related skills'): the explicit owns/delegates sentence resolves the overlap risk, so 5 fits.