Conciseness

The body is dense with domain-specific material (confidence table, sample-bias checks, ICP source table, templates) with only minor trimmable parts: the 'You are an expert customer researcher...' preamble, Mode 3 being described twice (in 'Three Modes of Research' and again in its own section), and a standard JTBD functional/emotional/social breakdown Claude already knows. It is efficient on the whole, so it sits above anchor 3 ('some unnecessary explanation') but short of anchor 5 ('every token earns its place') because of that duplication.

Actionability

For an instruction-only skill the guidance is fully executable: 'Extract: pains, triggers, desired outcomes, language used, objections, alternatives considered', a copy-paste synthesis template and persona template, and concrete numbers throughout ('$50/call, $5/survey; aim for 10 calls', '40% "very disappointed" benchmark', 'fewer than 5 independent data points per segment'). Not 4: there are no material gaps — templates and criteria cover the common cases.

Workflow Clarity

The sequence is explicit — 'Establish which mode(s) apply before proceeding', 'Mine what's already public (Mode 2) before you ask (Mode 3)', 'lead with #1 and #2' for questions, 'Ask the user which deliverable(s) they need before generating output' — with quality checkpoints ('Label every insight with a confidence level before presenting it', 'Flag contradictions'). Not 5: the checkpoints are quality guardrails rather than explicit validate→fix→retry loops, and the persona-generation stage is only loosely tied into the deliverable-selection flow; not 3 because sequence and checkpoints are both present and explicit.

Progressive Disclosure

The body keeps the overview and pushes detail to two well-signaled, one-level-deep references — 'read `references/interviews-and-surveys.md`' (with a bulleted summary of its contents) and 'read `references/source-guides.md` for detailed playbooks, search operators, and per-platform extraction tips' — and both files exist in the bundle and cover exactly what is claimed (verified against their headings). Not 4: the references are clearly signaled with content summaries and the split is appropriate, matching anchor 5.

Specificity

It lists several specific actions — 'conduct, analyze, or synthesize customer research' and 'analyzing existing research assets, mining online sources, AND running primary research (interviews and surveys)' — covering all three modes with minor gaps. Not 3 (well more than 1-2 actions, comprehensive across modes); not 5 because the verbs stay at a higher altitude than anchor-5 concreteness ('conduct customer research' borders on domain-generic).

Completeness

Both halves are explicit: the what ('conduct, analyze, or synthesize customer research... analyzing existing research assets, mining online sources, AND running primary research (interviews and surveys)') and the when ('Use when the user mentions...' followed by ~30 concrete trigger phrases). It matches anchor 5 exactly, including the explicit 'Use when' clause, so it is not capped at 3 and sits above anchor 4's 'when could be more explicit'.

Trigger Term Quality

The trigger list is comprehensive and natural — 'customer research,' 'ICP research,' 'talk to customers,' 'analyze transcripts,' 'voice of customer,' 'VOC,' 'jobs to be done,' 'JTBD,' 'Reddit mining,' 'G2 reviews,' 'PMF survey,' 'Sales Safari,' 'find out why customers churn/convert/buy' — including acronyms, full questions users would actually type, and named methodologies. Not 4: no common synonym or variation is missing.

Distinctiveness Conflict Risk

A clear niche with distinct triggers (VOC, PMF survey, Sales Safari, JTBD) plus explicit boundary guidance — 'For writing copy informed by research, see copywriting. For acting on research to improve pages, see cro.' Not 4: the disambiguation against the closest adjacent skills actively minimizes overlap risk, matching anchor 5.