Content
81%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 highly actionable, well-sequenced workflow body with exact commands, parse guidance, and strong validation feedback loops. Its main weakness is redundancy: the planning boundary and --store stickiness rules are each stated multiple times, inflating token cost without adding guidance.
Suggestions
State the planning-only boundary once (either the opening paragraph or the Guardrails) and cross-reference it from the other location instead of repeating it nearly verbatim.
Explain --store stickiness once in the Store selection paragraph and have steps 2 and 3 simply say "append the sticky --store flag" rather than re-describing the rule.
Merge the "re-read dependency artifacts from disk" instruction (currently in both step 5a and Guardrails) into a single place.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly procedural with no explanations of concepts Claude already knows, but noticeably repetitive: the planning boundary is stated in the opening paragraph and again almost verbatim in Guardrails; --store handling is explained in the Store selection paragraph, then repeated in steps 2 and 3; "re-read dependencies from disk, not conversation memory" appears twice. Fits the "mostly efficient but could be tightened" anchor rather than the "minor instances" of score 4. | 3 / 5 |
Actionability | Fully executable throughout: exact commands ("openspec new change \"<name>\" --schema \"<schema-name>\"", "openspec instructions <artifact-id> --change \"<name>\" --json"), the JSON fields to parse (applyRequires, requires edges, resolvedOutputPath), and per-artifact verification. Copy-paste ready and covers the common cases including store selection and skipped artifacts. | 5 / 5 |
Workflow Clarity | A clear numbered sequence with explicit validation checkpoints (re-run "openspec status --change \"<name>\" --json" after each artifact, verify the file exists at resolvedOutputPath), error-recovery paths (skipped/blocked artifacts, unclear input, existing change name), and a final status step. Matches the top anchor's feedback loops and checkpoint structure. | 5 / 5 |
Progressive Disclosure | No bundle files exist, so everything lives in one SKILL.md; that is appropriate for a cohesive multi-step workflow, and sections (Store selection, Steps, Artifact Creation Guidelines, Guardrails) are clearly headed and easy to navigate. Not 5 because some inline material (the long store-selection rules and duplicated boundary statements) could be consolidated or moved out, leaving minor organization gaps. | 4 / 5 |
Total | 17 / 20 Passed |