Conciseness

The body is dense with ast-grep-specific knowledge (silent --json/--update-all conflict, regex-misuse table, two-pass writes) that Claude does not already know, and it avoids explaining generic concepts. Minor over-explanation: the same invariants (pattern-is-not-regex, dry-run-first, two-pass) appear both in the 'Three things' section and again in 'Invariants', and the OmO MCP section is environment-specific padding for most users. Not 5 due to this redundancy; well above the verbose anchors.

Actionability

Every section is copy-paste executable: full helper invocations with flags and expected exit codes ('validate' → 'exit 2: regex \w not supported'), direct sg idioms, jq pipelines, stdin patterns, and --debug-query diagnostics. Specific examples cover the common search/replace/scan/validate cases exactly as the top anchor requires.

Workflow Clarity

The rewrite workflow is an explicit 5-step sequence with validation checkpoints and a feedback loop (search → dry-run → inspect preview → refine if wrong → apply only when right, 'Never apply a rewrite that you have not first dry-run'), and the 0-match section adds a prioritized error-recovery loop with 'Do not blindly retry'. For a batch/destructive operation skill, validation is central and complete, matching the top anchor with checklists (Invariants).

Progressive Disclosure

Clear overview with seven one-level-deep reference files, all of which exist in the bundle, each signaled with when-to-read guidance in 'Required reading (in order of priority)'. The gap: the body claims the skill ships install.sh (POSIX) and install.ps1 (Windows), but neither file exists in the bundle (no assets/ directory, scripts/ contains only ast_grep_helper.py), leaving helper 'install' as a dead-end reference — a minor organization flaw that keeps it below the top anchor.

Specificity

"Searches and rewrites code by AST shape" names two core concrete actions, and the when-clause adds "a codemod, a YAML rule" as further capability signals — several specific actions with minor gaps (e.g., scan/lint rule enforcement is only implied via "YAML rule"). Not 5 because the what-clause itself states only two actions rather than comprehensive coverage; not 3 because codemod and YAML-rule triggers go beyond 1-2 generic actions.

Completeness

Explicitly answers what ("Searches and rewrites code by AST shape across 25 languages") and when ("Use when the target is a syntax pattern (every call/class/import shaped like X, a codemod, a YAML rule) rather than literal text") with concrete trigger phrases, plus a negative boundary ("for plain strings, comments, or filenames, use rg"). Matches the anchor for a clear, explicit what-and-when with concrete triggers.

Trigger Term Quality

Natural trigger phrases are present: "syntax pattern", "every call/class/import shaped like X", "a codemod", "a YAML rule" — terms users would plausibly say. A few common synonyms are missing ("refactor", "find and replace", "AST"), keeping it below the comprehensive-synonym anchor of 5; well above the sparse keyword anchors of 2-3.

Distinctiveness Conflict Risk

Clear niche (AST-shape matching) with distinct triggers and an explicit disambiguation against rg for text-shaped queries, minimizing conflict risk with grep/search skills. Could not overlap with unrelated skills; matches the clear-niche anchor.