Conciseness

The body is largely lean and assumes Claude's competence — tool signatures like `firecrawl_search(query: "<sub-question keywords>", limit: 8)` and a compact report template with no filler explaining what research or MCP is. Not a 5 because of minor trimmable material: the drift warning is a bit padded and the "Examples" section of user prompts largely duplicates the description's trigger list; not a 3 because there are no padded explanations of concepts Claude already knows.

Actionability

Gives concrete, near-executable guidance: exact MCP tool calls with parameters (`web_search_advanced_exa(query: "<keywords>", numResults: 5, startPublishedDate: "2025-01-01")`, `crawling_exa(url: "<url>", tokensNum: 5000)`), concrete numeric targets ("2-3 keyword variations", "15-30 unique sources", "Read 3-5 key sources"), and a full report template. Not a 5 because the subagent section is outline-style ("Agent 1: Research sub-questions 1-2") rather than copy-paste-ready prompts, and tool calls are signatures rather than literal invocations; not a 3 because the guidance is far beyond pseudocode.

Workflow Clarity

A clear six-step sequence with well-defined steps, branch handling ("If the user says 'just research it' — skip ahead with reasonable defaults"), and quality checkpoints ("If only one source says it, flag it as unverified", "Acknowledge gaps"). Not a 5 because there is no explicit final validation pass before delivery (e.g., "verify every claim has a citation before writing the report") — the Quality Rules are stated as rules rather than as a step in the sequence; not a 3 because the sequence is explicit with most checkpoints present.

Progressive Disclosure

The single SKILL.md file (~150 lines) is well-organized into clearly signaled sections (When to Activate, MCP Requirements, Workflow, Parallel Research, Quality Rules) with no bundle files and no nested references. Not a 5 because at this length a couple of sections (the report template, the examples) could be split into reference files to keep the overview lean; not a 3 because all content is appropriately placed for a self-contained skill, with nothing buried or misfiled.

Specificity

Quotes several concrete actions — "Searches the web, synthesizes findings, and delivers cited reports with source attribution" — and names the concrete tooling ("using firecrawl and exa MCPs"). Not a 5 because coverage has minor gaps: no mention of sub-question planning, source deep-reading, or report structure; not a 3 because it lists several specific actions rather than just 1-2.

Completeness

Explicitly answers both parts: what — "Searches the web, synthesizes findings, and delivers cited reports with source attribution" — and when — "Use when the user wants thorough research on any topic with evidence and citations". This matches the anchor-5 pattern of a clear what paired with a concrete explicit trigger clause; not a 4 because the when-clause is already explicit and specific rather than merely serviceable.

Trigger Term Quality

Includes the natural phrases users would say — "thorough research on any topic", "evidence and citations" — which is good keyword coverage. Not a 5 because common variations like "deep dive", "investigate", or "look into" are missing (they appear only in the body, not the description); not a 3 because the primary natural term plus context phrases are present, not just bare keywords.

Distinctiveness Conflict Risk

The niche is mostly distinct — "Multi-source deep research using firecrawl and exa MCPs ... cited reports with source attribution" distinguishes it from single-source web-search or summarization skills. Not a 5 because "research on any topic" is broad and could overlap with a general web-search skill; not a 3 because the MCP tooling and citation/report framing give it a clear identity.