Conciseness

The body is lean and assumes Claude's competence — no padding or explanation of known concepts, and sections like 'Anti-patterns' carry real operational value. It falls short of 5 only because the 'read the real source, not the graph alone' guidance is repeated in the intro, workflow step 3, and the 'When NOT to use' section and could be trimmed to one place.

Actionability

Fully concrete, executable guidance: exact tool signatures with example arguments ('hivemind_graph_search({ pattern: "<symbol>" })', 'hivemind_graph_neighborhood({ file: "src/hooks/capture.ts" })'), the multi-token 'auth+handler' syntax, and the precise config key 'plugins.entries.hivemind.config.tuning.HIVEMIND_GRAPH_CWD'. Specific examples cover the common cases, matching the top anchor.

Workflow Clarity

The three-step workflow is a clear conditional decision tree ('Broad structural question?', 'Know the file?', 'Need the actual code?') with retry guidance ('try multiple keywords if the first misses') acting as a feedback loop. The skill is read-only querying, so no destructive/batch validation cap applies, and staleness caveats are explicitly flagged.

Progressive Disclosure

The skill is a short (~55 line), single-purpose overview with no external references needed and no bundle files, and content is organized into well-signaled sections (When to use, Tools, Workflow, When NOT to use, Anti-patterns). This satisfies the simple-skill exception: well-organized sections alone merit the top score.

Specificity

The description lists multiple concrete actions with comprehensive coverage: 'Query the local AST-derived code graph (functions, classes, calls, imports)' with enumerated question types ('what calls X, what does Y import, where is Z defined, blast radius of a change') and named tools ('hivemind_graph_search and hivemind_graph_neighborhood'). It matches the anchor 'lists multiple specific concrete actions; comprehensive coverage' and is not below it since nothing is vague or generic.

Completeness

The 'what' is explicit (query the AST-derived code graph) and the 'when' is present via concrete trigger questions ('for structural codebase questions — what calls X... blast radius of a change'), so it is not capped at 3 — equivalent trigger guidance exists. It does not reach 5 because there is no explicit 'Use when...' clause; the when-guidance could be more direct.

Trigger Term Quality

Good natural-phrase coverage — 'what calls X', 'what does Y import', 'where is Z defined', 'blast radius of a change' are phrases users actually say. It falls short of the 5 anchor because common synonyms like 'find usages', 'callers', 'references', or 'dependencies' are missing, but it is well above the 3 anchor ('some relevant keywords but missing common variations') since multiple natural trigger phrases are present.

Distinctiveness Conflict Risk

A clear niche (AST-derived code graph) with distinct tool names gives minimal conflict risk, but generic structural phrases like 'where is Z defined' or 'what does Y import' could also trigger general grep/symbol-search skills, matching 'mostly distinct; minor overlap risk' rather than the fully distinct 5 anchor.