Content
60%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.
The body is well-structured with a clear pre-change/post-change workflow and sensible routing, but it is held back by redundant sections (two 'when' lists, two skill-routing lists) and reliance on an undefined artifact vocabulary that leaves the guidance short of fully executable. Instruction-only format itself is not penalized.
Suggestions
Merge 'When to Activate' and 'When to Use' into one section — they currently repeat the same triggers almost verbatim, and consolidate 'Related Skills' with 'Routing Rules' to remove the second overlap.
Define or reference the project artifact terms (MAP, STATUS, LOG, CONTEXT, ADR, CONTRACT, TESTS, TEST-IDs, REGRESSION, Spec/Issue) in a short glossary or reference file; they are used as load-bearing concepts but never defined.
Add one or two concrete evidence-gathering examples, e.g. the specific `git grep`/`git diff` invocations or call-graph commands used to prove callers, so the 'prefer evidence from real imports and calls' instruction becomes executable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | "When to Activate" and "When to Use" repeat the same triggers nearly verbatim, and "Related Skills" plus "Routing Rules" cover the same four skills twice; the rest of the prose is tight, so this is noticeable redundancy rather than severe padding. | 3 / 5 |
Actionability | The ordered checklists ("Which entry points, direct callers, downstream consumers...") and routing rules with blockers are concrete, but the guidance hinges on undefined artifact jargon (MAP, STATUS, LOG, CONTEXT, TEST-IDs, Spec/Issue) with no bundle files or glossary, and only names `git grep`/`git diff` without executable examples. | 3 / 5 |
Workflow Clarity | A clear sequence runs from "Before the Change" (1–6) through routing rules with an explicit blocker ("Treat any missing item as a blocker; do not execute rollback") to "Post-Implementation Alignment" (1–5), which acts as a scope-drift feedback loop; checkpoints are present but mostly implicit question lists rather than explicit validate-fix-retry steps. | 4 / 5 |
Progressive Disclosure | No bundle files exist and the ~80-line body is appropriately self-contained with clear headers; minor organization gaps (the duplicated trigger sections and undefined artifact ecosystem) keep it below a 5. | 4 / 5 |
Total | 14 / 20 Passed |