Content
67%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 a well-structured, actionable instruction skill: numbered checks with concrete bash commands, a DON'T list, and a completion checklist give clear execution guidance. Its main weakness is token efficiency — a concept re-explanation and session-specific sections add tokens without reusable value, and one code example (the ast-grep dead-code search) is not actually executable.
Suggestions
Remove the "Pattern" concept explanation (Claude already knows that untested/unwired code isn't complete) and the session-specific "Source Sessions" section; the DO/DON'T/checklist already carry the guidance.
Fix or replace the orphaned-implementation search: the `ast-grep --pattern 'async function $NAME() { $$$ }' | xargs -I {} grep -r "{}" src/` pipe won't work because ast-grep emits `file:line:match` lines, not bare names — use a targeted, tested command or drop the step.
Generalize the end-to-end step's placeholder (`uv run python -m my_feature`) into a concrete, adaptable recipe, and add a brief "if the check fails" recovery note to close the workflow-clarity gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The "Pattern" section re-explains a concept Claude already knows ("Infrastructure is not done when the code is written - it's done when it's wired... Dead code... is wasted effort"), and the "Source Sessions" section ("This session: Architecture gap discovery - DAG built but not wired...") is session-specific clutter with no reusable value. This is more than the "minor instances" of anchor 4, but the DO/DON'T/checklist core is efficient, so it sits at anchor 3. | 3 / 5 |
Actionability | Steps ship concrete, mostly executable commands (e.g., `grep -r "claude -p" src/`, `ls -la .claude/hooks/my-hook.sh`, the debug-log then trigger then `cat /tmp/debug.log` loop). Minor gaps: the `ast-grep --pattern ... | xargs -I {} grep -r "{}" src/` pipe is not sound as written (ast-grep emits `file:line:match`, not bare function names), and `uv run python -m my_feature` is a placeholder, so it stops short of anchor 5's copy-paste-ready coverage. | 4 / 5 |
Workflow Clarity | Five clearly sequenced numbered steps plus an explicit validation mechanism — the "Completion Checklist" ("Before declaring infrastructure complete" with checkboxes) and a step-4 end-to-end verification. Not a destructive/batch skill, so no cap applies; it falls short of anchor 5 only because there is no error-recovery/feedback-loop guidance for what to do when a check fails. | 4 / 5 |
Progressive Disclosure | A single-file skill with no bundle files and well-organized sections (Pattern / DO / DON'T / Checklist / Example) with no nested references. Minor placement gaps — the project-specific "Example: DAG Task Graph" and the "Source Sessions" section are session artifacts that dilute the general overview, keeping it at anchor 4 rather than 5. | 4 / 5 |
Total | 15 / 20 Passed |