Content
88%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.
An exemplary instruction-only skill: copy-paste-ready commands, a validated workflow with explicit gates, and hard-won project-specific judgment (the six traps) that Claude could not reconstruct on its own. The only trimmable fat is slight narrative padding around the historical failures, and the long inline traps section is a marginal disclosure opportunity.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and earns nearly every token: commands, project-specific data (median 3.6M vs mean 15.4M), and failure cases Claude could not know. It sits at 4 rather than 5 because the two-wrong-findings narrative in "Why the order below matters" and a handful of long rhetorical sentences (e.g., "offering the answerable ones as though they were the whole picture is how a partial view becomes a wrong one") could be tightened without losing content. | 4 / 5 |
Actionability | Guidance is fully executable: copy-paste-ready commands (`bun run --cwd tools/pr-metrics report -- --coverage`, the `gh pr diff … | grep -E` dependency-bump check), concrete flag semantics (`--json` carries notes, `--split-at <commit>`), and a precise output contract (number, sample, exclusions, what would overturn it). This matches the 5 anchor for an instruction-only skill per the code_vs_instruction scoring note. | 5 / 5 |
Workflow Clarity | The sequence is explicit and justified (Step 0 coverage → Step 1 run what you need → Step 2 report) with real validation gates: "write them down before continuing", "If nothing is rated, say so at the top of your answer", per-claim exclusions and falsifiers, and a hard rule against unsupported recommendations. These are explicit checkpoints with error-recovery guidance, matching the 5 anchor; the 4 anchor's "minor validation gaps" is not in evidence. | 5 / 5 |
Progressive Disclosure | No bundle files exist, so this scores on body structure: clear section headers, well-signaled one-level pointers to external material (`tools/pr-metrics/queries.sql`, `wiki/conventions/session-metrics.md`, `tools/pr-metrics/README.md` in "What consumes this"), and appropriately routed handoffs to sibling skills. It stops short of 5 because at ~139 lines everything is inlined — the six-traps section (~60 lines) is arguably reference material — and the skill exceeds the under-50-lines exception for a sections-only 5. | 4 / 5 |
Total | 18 / 20 Passed |