The body is mostly efficient expert reference material (CRAAP scoring, verification levels, search operators) but includes unnecessary explanation Claude already knows, e.g. 'Confirmation bias: Favoring information that confirms your initial hypothesis' and 'Stack Overflow answers may be outdated'; it is not 3 because of this padding, and not 1 because the bulk is genuinely additive, not boilerplate.

It provides copy-paste-ready templates (narrative synthesis skeleton, structured synthesis, APA/footnote/numbered citation formats, brief and detailed report templates) and a concrete CRAAP scoring rubric with search operators; it is not 2 because guidance is specific and complete rather than pseudocode, and the instruction-only nature is fully actionable per the scoring notes.

The 5-phase process (Define, Search, Evaluate, Synthesize, Verify) is clearly sequenced with a Verify checkpoint, and contradiction resolution plus gap analysis act as feedback loops; it is not 2 because checkpoints are explicit, and not below 2 because the multi-step process is unambiguous.

Content is well-organized into sections but is a 328-line monolithic SKILL.md with no bundle files and reference-worthy material (citation formats, output templates, domain tips) kept inline; it is not 3 because the file exceeds the 50-line simple-skill threshold and content that should be split is inline, and not 1 because sections are clearly labeled rather than a nested-reference mess.

The description names six concrete capabilities — 'methodology, source evaluation, search optimization, cross-referencing, synthesis, and citation formats' — matching the anchor for multiple specific concrete actions; it is not below 2 because it is comprehensive, and the actions are too explicitly listed to be merely domain-plus-some.

It clearly states what the skill does but has no 'Use when...' clause or equivalent explicit trigger guidance, which per the judging guidelines caps completeness at 2; it is not 3 because the 'when' is entirely absent, and not 1 because the 'what' is explicit and detailed.

Terms like 'deep research', 'source evaluation', and 'search optimization' are relevant but miss common phrasings a user would actually say ('research this', 'verify these claims', 'find reliable sources'); it is not 3 because natural keyword coverage is partial, and not 1 because the keywords present are reasonable rather than jargon.

'Expert knowledge for AI deep research' carves a recognizable niche but 'research' is broad and could overlap with general web-search or research skills; it is not 3 because the triggers are not distinct enough to rule out conflicts, and not 1 because it is more specific than generic 'helps with research'.