Conciseness

The body is dense and directive with zero padding: "Fast by default. Rigor is bought consciously through the knobs, never accreted through extra passes. One gate, light checkpoints, no ceremony." No concept Claude already knows is explained; every line instructs ("A claim is a sentence with a source. Publisher, publication date, access date. No naked numbers."). Matches the 5 anchor ("lean and efficient; assumes Claude's competence; every token earns its place"); 4 would require identifiable over-explanation to trim, and none is present.

Actionability

Copy-paste-ready commands with exact arguments: "uv run {project-root}/_bmad/scripts/memlog.py" with --type values enumerated, the resolve_customization.py and resolve_config.py invocations with --skill/--key flags, "npx skills add bmad-code-org/BMAD-METHOD --skill bmad", and "uv run {skill-root}/scripts/recon_kit.py tally {doc_workspace}/.memlog.md" for claim counts (never hand-counted). Concrete routing table, concrete failure paths, concrete JSON end-state. Matches the 5 anchor; not 4 because the executable commands present are complete rather than having minor gaps.

Workflow Clarity

The activation sequence is exemplary — numbered steps with explicit error-recovery feedback loops ("Script not found: BMad is not set up here. Offer to run the bmad skill's setup... then run the command again") and a named hard stop ("hold the plan gate — the one hard stop"), plus a resume-from-disk guarantee. But the core research loop's validation checkpoints are delegated rather than stated — "then run the loop" points at references/run.md and verification.md — so the body itself has minor checkpoint gaps, fitting anchor 4 rather than 5's fully-inline validation steps and checklists.

Progressive Disclosure

A textbook router: each intent in the table names exactly which files to load ("references/run.md, then references/verification.md + references/synthesis.md"), and every referenced file exists in the bundle (verified: draft.md, process.md, run.md, verification.md, synthesis.md, lifecycle.md, codebase.md, selection.md, finalize.md, plus scripts/recon_kit.py and assets/ templates). One level deep, clearly signaled, load-only-what-it-names. Matches the 5 anchor; no buried or nested references.

Specificity

The capability statement is a single abstract clause — "Decision-grade research on a topic or this repository" — and the only concrete actions ("draft a research prompt", "process this research report") appear as trigger quotes rather than capability descriptions; it never names the draft/process/run/codebase services it performs. Matches anchor 3 ("names domain and 1-2 concrete actions, but not comprehensive"): not 4 because it does not list several specific actions; not 2 because it goes beyond naming the domain with a couple of concrete action phrases.

Completeness

Both halves are explicit: the "what" ("Decision-grade research on a topic or this repository") and a maximal "when" — "Use when the user says" plus 14 concrete trigger phrases and a negative boundary ("Not for a quick code question answered in chat"). Matches the 5 anchor (both what and when with concrete trigger phrases); 4 would require the 'when' to be less explicit, which it is not.

Trigger Term Quality

Fourteen natural phrases users would actually say — "deep recon", "research this", "market research", "what users say about X", "literature review", "help me choose between", "why is X built this way" — covering synonyms, question forms, and codebase variants. Clearly the comprehensive-coverage anchor (5); no common variation is obviously missing, so 4's "a few natural terms missing" does not fit.

Distinctiveness Conflict Risk

A clear research niche with mostly distinct triggers, sharpened by the exclusion clause "Not for a quick code question answered in chat"; but "research the codebase", "research how X works here", and "literature review" could also plausibly route to general web-search or code-explainer skills. Fits anchor 4 ("mostly distinct; minor overlap risk with closely related skills"); not 5 because of that residual overlap with closely related skills, not 3 because the negative boundary and multi-trigger phrasing keep conflicts minor.