Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is actionable and well-sequenced for a stance skill, with concrete commands, file paths, and a capture table. It loses points on conciseness due to repeated stance restatements and on progressive disclosure since it is a single long inline document with no reference layering.
Suggestions
Consolidate the redundant stance statements ("The Stance", "What You Don't Have To Do", "Guardrails") into one section to cut repetition.
Move the four long ASCII dialogue examples into a reference file (e.g. EXAMPLES.md) and keep one short illustrative example inline.
Tighten or trim the ASCII art boxes, which consume significant tokens for largely stylistic effect.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and assumption-respecting, but the stance is restated across "The Stance", "What You Don't Have To Do", and "Guardrails", and the long ASCII dialog blocks pad the token budget beyond what the core guidance needs. | 2 / 3 |
Actionability | Provides concrete, executable guidance for an instruction-only skill: the `openspec list --json` command, specific artifact paths, an insight-to-artifact capture table, and example offer phrasings. | 3 / 3 |
Workflow Clarity | Deliberately a stance rather than a workflow, yet the "When a change exists" section gives a clear 1–4 numbered sequence; no validation checkpoint is needed for a non-destructive thinking activity, and the single core action is unambiguous. | 3 / 3 |
Progressive Disclosure | Well-organized into labeled sections but monolithic: ~275 lines all inline with no external references or progressive layering, and the illustrative dialogues are not split out despite the file's length. | 2 / 3 |
Total | 10 / 12 Passed |