Conciseness

Noticeably verbose: citation rules appear in four sections (Contract, Output Format, Quality Rules, Citation in Answers), source precedence in three, and "never hallucinate" in four; the ~55-line Token-Budget Awareness section inlines protocol minutiae (cost formulas, fractional-budget compatibility, adoption verification) that belongs in the referenced protocol doc. Not 3 because the duplication and inlined protocol detail go beyond 'some unnecessary explanation'; not 1 because most content is genuinely non-obvious operational guidance rather than concepts Claude already knows.

Actionability

Fully executable, copy-paste-ready commands with realistic arguments cover the common cases: `gbrain query "<q>" --since 2026-09-01 --until 2026-09-30`, `gbrain graph-query companies/acme --type works_at --direction in`, `gbrain recall --query 'zebra telescope' --budget-tokens 75 --budget-policy query_first --json`, and `--return-unit page --token-budget 6000`, plus a complete MCP JSON request example.

Workflow Clarity

Five clearly sequenced phases with explicit validation and error-recovery feedback loops: step 5 mandates reading notices before answering, and 'When it fails' prescribes concrete retries (try display name/synonyms/hybrid before declaring a gap; page or narrow before claiming absence; search by title before reporting a page missing). Matches the anchor for clear sequence with explicit validation steps and feedback loops.

Progressive Disclosure

Section headers are clear, but the skill ships no bundle files while referencing external docs (docs/protocol/MEMORY_VERBS_v1.md, ../conventions/agent-operator-protocol.md, skills/conventions/brain-first.md) and inlining the token-budget protocol detail inline instead of splitting it out. Not 4 because inlining ~55 lines of protocol-level content is more than a minor organization gap; not 2 because the 13-section structure is real and references are clearly signaled.

Specificity

Lists several concrete actions — "3-layer search, synthesis, and citation propagation" — grounded in a specific domain ("the brain's knowledge"). Not a 5 because the actions are terse and unexplained (the three layers are never named), leaving minor gaps in coverage.

Completeness

Explicitly answers both what ("Answer questions using the brain's knowledge with 3-layer search, synthesis, and citation propagation") and when ("Use when the user asks a question, wants a lookup, or needs information from the brain"). Not 5 because the when-clause triggers are broad abstractions rather than concrete trigger phrases users would say.

Trigger Term Quality

"asks a question", "wants a lookup", "needs information" are natural phrases but generic, and common variations (e.g., "tell me about", "who is", "search for" — which exist only in the separate triggers field, not the description) are missing. Not 4 because the coverage of natural terms is thin, not just missing 'a few'.

Distinctiveness Conflict Risk

The what-clause carves a niche (brain knowledge, citations), but "Use when the user asks a question... or needs information" would match virtually any question or information request, creating real overlap risk with general search/research skills. Not 4 because the trigger scope is broad rather than a minor overlap with closely related skills.