Content
72%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A highly actionable, well-structured workflow body with concrete commands and exact JSON field guidance. Its main costs are repetitive Guardrails restating the Steps and the absence of a post-write validation loop for multi-artifact edits.
Suggestions
Collapse the Guardrails section into only the rules not already stated in the Steps (e.g. drop the repeated "NEVER edit code", "confirm every edit", and "existingOutputPaths vs glob" lines) to remove the conciseness redundancy.
Add an explicit post-write verification step — e.g. re-run `openspec status --change "<name>" --json` or `openspec validate` after edits — to close the feedback loop for batch artifact revisions.
Consider moving the "Store selection" paragraph and the sibling-skill pointer map (step 6 + the core-profile fallback) into a short reference section or bundled file to keep the main workflow sequence leaner.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and assumes Claude's competence, but the Guardrails section largely re-states rules already in the Steps ("NEVER edit implementation code", "no new artifacts", "confirm every edit", "never branch on hardcoded artifact names", "existingOutputPaths" vs glob) — a tighten-able redundancy. | 2 / 3 |
Actionability | Provides copy-paste-ready commands (openspec list/status/instructions/store with --json), names the exact JSON fields to read (schemaName, artifacts, isComplete, existingOutputPaths), and specifies which paths to edit vs. avoid — fully executable guidance. | 3 / 3 |
Workflow Clarity | Six clearly sequenced steps with confirm-before-write checkpoints and rejection handling, but for batch edits across multiple artifacts there is no post-write verification/re-validation step (e.g. re-running openspec validate), which caps clarity per the destructive/batch feedback-loop guidance. | 2 / 3 |
Progressive Disclosure | Well-organized into labeled sections (Store selection, Input, Steps, Output, Guardrails) with one-level-deep, clearly signaled cross-references to sibling skills and CLI commands; no nested references and no monolithic reference wall, appropriate for a self-contained workflow skill with no bundle files. | 3 / 3 |
Total | 10 / 12 Passed |