Content
73%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.
A strongly structured verification workflow with copy-ready CLI commands, explicit conditionals, graceful degradation for missing artifacts, and a severity-gated archive decision. The main cost is token weight: repeated issue templates and a few explanations of things Claude already knows could be consolidated to tighten the body.
Suggestions
Define the issue format once (e.g., "Every issue: severity, message, and a Recommendation with file/line references where applicable") and list the six checks as one-line entries, removing the repeated issue-plus-Recommendation pattern in steps 5-7.
Delete explanations of knowledge Claude already has, such as the checkbox syntax ("`- [ ]` (incomplete) vs `- [x]` (complete)") and "Count complete vs total tasks".
Extract the full report template and issue message examples into a references/ file (e.g., REPORT_FORMAT.md) and keep a one-line pointer in SKILL.md, slimming the main body below its current ~157 lines.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient prescriptive guidance, but includes unnecessary explanation Claude already knows ("Parse checkboxes: `- [ ]` (incomplete) vs `- [x]` (complete)", "Count complete vs total tasks") and repeats the issue-plus-"Recommendation:" template six times across steps 5-7 instead of defining the issue format once. This fits anchor 3 (mostly efficient but could be tightened) rather than 4, where over-explanation would be only minor. | 3 / 5 |
Actionability | Concrete, executable commands are given throughout ("openspec list --json", "openspec status --change \"<name>\" --json", "openspec instructions apply --change \"<name>\" --json") plus an explicit report template and file-reference format. It falls short of anchor 5 because the codebase-verification steps remain high-level ("Search codebase for keywords related to the requirement") with no worked example of a filled-in report. | 4 / 5 |
Workflow Clarity | Steps 1-8 form a clear numbered sequence with explicit decision points ("If no change name provided, prompt for selection", "If no design.md: Skip... and note"), a severity-gated final assessment that explicitly controls archiving ("Fix before archiving" / "Ready for archive"), and error-recovery rules ("When uncertain, prefer SUGGESTION over WARNING", WARNING over CRITICAL). This matches anchor 5: explicit validation steps, feedback rules, and a checklist-like three-dimension report; the operation is read-only verification, so no destructive-operation cap applies. | 5 / 5 |
Progressive Disclosure | The bundle contains no references/ and the body is well-organized into Steps, Verification Heuristics, Graceful Degradation, and Output Format sections with all content operational rather than misfiled reference material. At ~157 lines it exceeds the under-50-line exception for a reference-free 5, and the report template and issue-message formats could arguably be extracted to a reference file, so anchor 4 (good structure, minor organization gaps) is the best fit. | 4 / 5 |
Total | 16 / 20 Passed |