Content
88%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 well-engineered workflow skill: exact commands, explicit pre-write validation, feedback loops, and firm guardrails against writing code or advancing the build frontier. Its weaknesses are mild repetition of guardrail statements, a few tangled sentences, and edge-case material that lengthens the file without a quick-path summary.
Suggestions
Consolidate the repeated guardrails (never write to resolvedOutputPath, leave skipped artifacts untouched, confirm before writing) into the Guardrails section only, and reference them from steps 2, 4, and 6 instead of restating them.
Add a 2-3 line quick-path summary (e.g., status -> read artifacts -> propose revisions -> confirm -> write) at the top so the common case is scannable before the store-selection and root-null edge cases.
Split the dense store-selection and project-check paragraphs into short labeled bullets to reduce re-reading cost per invocation.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Nearly every token carries non-obvious operational detail (store flag stickiness, "root": null semantics, existingOutputPaths vs resolvedOutputPath) with no padding or explanation of known concepts, matching anchor 4. Not 5: several guardrails are repeated across steps 2, 4, and the Guardrails section, and some sentences (e.g., the store-selection paragraph) are convoluted and could be tightened. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands with exact syntax (openspec status --change "<name>" --json, openspec instructions "<artifact-id>" --change "<name>" --json, openspec list --json), concrete JSON fields to read (root, existingOutputPaths, lastModified), and specific path-creation checks. This is an instruction-only skill and its guidance is concretely actionable, so absence of code is not penalized. | 5 / 5 |
Workflow Clarity | Six clearly sequenced steps with explicit validation checkpoints (project-root check before any write, re-refreshing status/instructions immediately before creating a glob file, confirm-before-write on every edit) and feedback loops for error recovery (stop on skipped artifacts, missing dependencies, or failed checks). Matches the anchor-5 pattern of validate-then-proceed with retry/stop guidance. | 5 / 5 |
Progressive Disclosure | Single-file skill with no bundle files; content is appropriately placed for a CLI-driving workflow and organized under clear bolded sections (Store selection, Project check, Input, Steps, Output, Guardrails) with a one-line overview. Anchor 4 rather than 5: at ~100 lines, the interleaved edge-case handling (store selection, root-null disambiguation, the 5-item glob procedure) makes the main happy path harder to scan, a minor organization gap. | 4 / 5 |
Total | 18 / 20 Passed |