Content
86%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
Dense, expert-level content where every token earns its place — the gotchas section in particular encodes knowledge that could not be reconstructed from the code alone. The main weakness is a small internal inconsistency: the corrected guard-clause Cypher omits the file-id anchoring that the skill's own unbounded-scan warning says is mandatory, slightly undercutting copy-paste reliability and leaving the advertised debugging flow without an explicit procedure.
Suggestions
Add file-id or symbol-span anchoring (e.g. a WHERE clause on the BasicBlock id-prefix) to the example Cypher so it is bounded as the skill's own first gotcha requires.
Add a short ordered procedure for the debugging use case, e.g.: empty result → check the RepoMeta.pdg.maxCdgEdgesPerFunction / maxReachingDefEdgesPerFunction probe → if no PDG layer, re-index with --pdg.
Clarify how the L1/L2/L5 build-order layers relate to when each mode becomes queryable (e.g. whether `controls` needs both M1 and M5 indexed).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line carries non-obvious, load-bearing detail (no rel-property index in LadybugDB, 1-based vs 0-based line offsets with the dual +1 rationale, binary switch-arm labels, the RFC #567 Cypher correction) with zero padding or explanation of concepts Claude already knows. | 5 / 5 |
Actionability | Concrete tool invocations for both modes, a real Cypher query, and exact probe fields (RepoMeta.pdg.maxCdgEdgesPerFunction) are given, but the flagship Cypher example is unanchored as written — which the skill's own first gotcha says is unbounded — so it is not safely copy-paste ready without the anchoring it only alludes to. | 4 / 5 |
Workflow Clarity | The body is coherently sequenced (When to Use → layered build order → the two modes → corrected Cypher → gotchas → extension guidance) and covers the empty-result error cause via the meta-probe note, but there is no explicit ordered procedure for the debugging use case it advertises ("why is the pdg_query result empty?"). | 4 / 5 |
Progressive Disclosure | A single-file skill with no references/, scripts/, or assets/ bundle and nothing that warrants splitting: the whole ~84-line surface is needed at once, headers are clear, and navigation is easy with no buried or nested references. | 5 / 5 |
Total | 18 / 20 Passed |