Conciseness

The body is dense and nearly filler-free: no explanations of concepts Claude already knows, and every section carries operational content (cheat sheet, decision flow, key priority chain, runtime steps). Minor trim opportunities keep it at 4 rather than 5: the vertical-domain rule is stated twice (Trigger section and Decision Flow), the two-CLI rationale ('on purpose (smaller review/maintenance surface)') and the base/latest/full Docker image detail are justification prose, and the Required-params rule duplicates what `doc` renders.

Actionability

Fully executable, copy-paste-ready commands cover the common cases: the cheat sheet shows concrete invocations for search (general, --tag, --domain/--sdp), get_sub_domains, batch_search (flag and JSON forms), and extract; the `<cmd>` placeholder is explicitly defined ('Replace <cmd> with the command from runtime.conf (for example, python3 <skill_dir>/scripts/anysearch_cli.py)'); runtime detection/persistence and .env setup are given as runnable commands; and failure paths are concrete ('run <cmd> <subcommand> --help rather than full doc') with invalid invocations enumerated ('extract --format markdown ... Invalid').

Workflow Clarity

A clear decision flow with explicit branches (Path 1 / HYBRID / Path 2), a mandatory get_sub_domains-before-vertical-search precondition, and a runtime bootstrap that validates before persisting (detect → entry-test `doc` → write runtime.conf, with 'replace it instead of appending'). Error-recovery feedback loops are present for the failure modes that matter: argument-schema failure → `--help`, service/quota/network failure → the Fallback rule ('inform the user and MAY fall back to other available search methods if the user approves'), which satisfies the batch/validation cap guidance. This is not 4 because checkpoints are explicit rather than implied at every branch.

Progressive Disclosure

Structure is good (Overview → Trigger → Cheat Sheet → Decision Flow → Key Config → Runtime → Security → Maintenance) and the full API reference is correctly pushed one level deep via the `doc` command (rendered from scripts/shared/doc_spec.md, which exists in the bundle). But scored against the actual bundle, several referenced files do not exist: 'See MAINTENANCE.md in this skill directory and docs/anysearch-skill.md', 'cp <skill_dir>/.env.example <skill_dir>/.env' (a command that fails — no .env.example is shipped), and 'see LICENSE and NOTICE in this skill directory' are all dangling references. Not 2, because the core content is well placed and the primary deep-reference mechanism (`doc`) works; not 4, because broken navigation is more than a minor organization gap.

Specificity

Lists multiple concrete actions with comprehensive coverage: 'live web and vertical-domain search, parallel batch search, and full-page URL extraction', plus per-command triggers ('call get_sub_domains before vertical search', 'use batch_search when the query is ambiguous or multi-intent', 'use extract to turn a page into Markdown') and auth options. This matches the score-5 anchor; nothing is generic, though the breadth is near the 4/5 boundary, every capability is named concretely.

Completeness

The 'what' is explicit (search, batch search, extract) and 'when' guidance exists per capability ('Prefer vertical search for domain-specific or realtime queries', 'use batch_search when the query is ambiguous or multi-intent'), so this clears the score-3 cap for missing trigger guidance. It sits below 5 because there is no single explicit overall activation clause ('Use when you need to search the web, fact-check, or read a page') — the when is distributed across features rather than stated as a concrete trigger phrase for the skill as a whole.

Trigger Term Quality

Good natural-keyword coverage: 'web search', 'realtime queries', 'extract', 'turn a page into Markdown', and a strong domain list ('finance, academic, security, legal, code, travel, health'). Falls short of the 5 anchor because it lacks synonyms users commonly say ('look up', 'find', 'browse', 'fact-check') and any equivalent of file-extension markers; above 3 because the terms present are the phrases users would naturally say.

Distinctiveness Conflict Risk

The 'via AnySearch' marker and the vertical-domain niche (Stock/CVE/DOI-style structured queries, get_sub_domains/batch_search vocabulary) give it a distinct identity, but the leading trigger space 'live web and vertical-domain search' overlaps heavily with built-in and generic web-search skills, matching the score-3 anchor ('somewhat specific but could still overlap with similar skills'). Not 4: the description does not delineate when to prefer it over ordinary web search in a way that would prevent wrong-skill triggering on a plain 'search the web' request.