Content
77%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.
The body is a well-sequenced, highly actionable verification workflow with strong validation checkpoints and feedback loops, but it is dense with inline policy prose and lacks a bundled reference structure to offload detail. Splitting decision/policy rules into reference files would improve progressive disclosure.
Suggestions
Move the dense policy/decision rules (e.g. the Step 1b uncertainty principles, the dirty-worktree attribution rules, the spec-drift A/B/C options) into reference files and keep SKILL.md as a sequenced overview with clearly signaled links.
Add the referenced `comet-classic/reference/*.md` files into a bundled `references/` directory so the progressive-disclosure structure is real rather than external-only.
Tighten the repeated "不得/禁止" caveats into a single consolidated constraints section to reduce token cost while preserving the rules.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and mostly efficient — concrete CLI commands and rules with no padding or basic-concept explanation — matching the 'efficient; minor instances of over-explanation that could be trimmed' anchor; not 5 because several policy caveats (repeated "不得/禁止" reminders) and verbose decision prose could be tightened. | 4 / 5 |
Actionability | It provides concrete, executable guidance — `comet state check`, `comet state transition <change-name> verify-fail`, `git diff --stat <base-ref>...HEAD`, `comet guard ... --apply` — covering common cases; not 5 because some flows are prose-heavy rather than scripted and use placeholders, leaving minor gaps. | 4 / 5 |
Workflow Clarity | Steps are clearly sequenced (0a→0b→1→1b→2→2a/2b→3→exit) with explicit validation checkpoints (`comet state check ... verify`, `comet guard ... --apply`) and a feedback loop (verify-fail → /comet-build → retry, capped at 3 failures plus idempotency notes), matching the 'clear sequence with explicit validation steps; feedback loops' anchor. | 5 / 5 |
Progressive Disclosure | It references external `comet-classic/reference/*.md` paths clearly by name but inlines substantial policy/decision logic that belongs in separate files, and no bundle files (references/scripts/assets) are present in the skill, matching the 'some structure but could be better organized; content that should be separate is inline' anchor; not 4 because of the heavy inlining and absence of a bundled reference structure. | 3 / 5 |
Total | 16 / 20 Passed |