Content
100%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.
An exemplary skill body: tight, imperative, and entirely actionable, with a real validation loop ('a skill that was never executed is a draft') and concrete tool-specific checklists. Nothing is padded and nothing Claude already knows is re-explained.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~35-line body is lean and imperative with zero concept explanations — 'Write `.factory/skills/verify-<repo>/` for the next agent, not as a human tutorial' sets the tone, and every subsequent line instructs rather than explains. Every token earns its place, matching the level-5 anchor. | 5 / 5 |
Actionability | Provides a copy-paste-ready generator command (`node <factory-supervise>/scripts/write-verification-skill.mjs --code-root ... --factory-root ...`) with explicitly justified placeholders, the exact evidence path `.factory/evidence/<ticket>/<task>/`, named required sections (Launch, Doctor, Drive, Evidence, Cleanup), and a precise checklist of surf subcommands. Fully executable guidance covering the common cases — the level-5 anchor. | 5 / 5 |
Workflow Clarity | Clear sequence — interview from checkout, generate files, replace placeholders, prove one feature, persist the rerunnable command — with an explicit validation feedback loop: 'Run the generated skill once... confirm the evidence file is still there. Fix the skill if that fails. A skill that was never executed is a draft.' Not a 4, because validation is explicit with a fix-and-retry loop rather than merely present. | 5 / 5 |
Progressive Disclosure | A single self-contained SKILL.md under 50 lines with no external references needed and clear sections (Interview, Prove one feature). Per the simple-skill guidance, well-organized sections alone warrant the top score; there is no bundle to disclose further. | 5 / 5 |
Total | 20 / 20 Passed |