Conciseness

Every line carries information Claude could not know: Exa mode table with result counts/text caps/timeouts, flag semantics ("--include-domain is a hard host filter, not a quality filter"), sidecar rules, and env-var/1Password handling. No known concepts are re-explained and no section is padded, matching 'Lean and efficient; every token earns its place'.

Actionability

The Running and Validation sections give copy-paste-ready commands with full paths covering report, query-file, json, validate, and doctor invocations, plus the mode table, overrides, and validate's output shape ("{ok, errors, warnings, mode, synthesis, queryCount}") and exit semantics. This matches 'Fully executable; copy-paste ready commands; specific examples cover the common cases'.

Workflow Clarity

The sequence is clear and validation is explicit: run the report (mode chosen up front), write findings.md plus sidecar, then "run `validate` and stop", with "exits 0 when there are no errors" and a manual read-back checklist (contradicted claims, off-topic sources, unsupported recommendations). It sits below 5 only because no fix-and-revalidate loop is spelled out for validation failures — 'Clear sequence with most checkpoints present; minor validation gaps'.

Progressive Disclosure

The body is a lean overview that pushes detail into the script: `scripts/deep-research` exists in the bundle and references (`agents/planner.md`, `templates/findings.md`, `../code-quality/SKILL.md`) are one level deep and each is signaled with its purpose. It falls short of 5 because `templates/findings.md` and `agents/planner.md` are referenced but not present in the bundle, a minor organization gap matching 'Good structure; most content is appropriately placed; references mostly clear'.

Specificity

The description names the domain ("research tasks, architectural investigations, and vendor, library, or technology comparisons") but lists no concrete actions — there is no verb stating what the skill actually does (e.g., produce a report, run a search). This matches the anchor 'Names the domain but actions are minimal or generic'; it is below 3 because no capability at all is stated, and above 1 because the domain categories are specific.

Completeness

Only the 'when' is present ("Load for research tasks, ...") with no 'what' — the description never says what the skill does (Exa-powered research producing an evidence-backed findings.md report, per the summary field). This matches the anchor 'only "when" is present without "what"'. It cannot be 3 because that anchor requires a clear 'what', and the explicit 'Load for...' trigger keeps it above 1.

Trigger Term Quality

Natural terms users would say are present: "research", "vendor, library, or technology comparisons", "architectural investigations". This matches 'Good keyword coverage; a few natural terms missing' — variants like "evaluate options", "X vs Y", "deep research", or "due diligence" are absent, keeping it below 5, while coverage clearly exceeds the 'some relevant keywords but missing common variations' level of 3.

Distinctiveness Conflict Risk

"vendor, library, or technology comparisons" is a fairly distinct niche, but "research tasks" and "architectural investigations" are broad and would overlap with other research or code-analysis skills. This matches 'Somewhat specific but could still overlap with similar skills' — above 2 because the comparison framing is not generic, below 4 because of the broad leading terms.