Content
67%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-structured, highly actionable coordinator skill with concrete commands, a gated execute path, and clear safety boundaries. Its main flaws are heavy repetition of the AskUserQuestion boundary contract and the absence of a post-apply verification checkpoint.
Suggestions
State the orchestrator/AskUserQuestion boundary once (e.g., in 'Key constraint') and reference it from elsewhere, cutting the repeated 'Canonical contract' lines and trimming the [AUTO] annotation to a sentence.
Add a post-apply verification step, e.g., re-run 'moai harness status' after 'apply --execute' to confirm the pending-proposal count decreased, completing the feedback loop for the destructive write.
Show a small example of the emitted payload JSON in Step 3 so the tool output format is copy-paste ready rather than described field-by-field.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly terse (command blocks, bullet schemas), but the orchestrator/AskUserQuestion boundary is restated at least four times (intro, 'Key constraint', Step 3, the L5 table row), the 'Canonical contract: .claude/rules/.../askuser-protocol.md' pointer is repeated three times, and the long [AUTO] annotation comment adds policy meta-text. This repetition could be tightened to a single statement, matching the 'some unnecessary explanation' anchor rather than 'minor instances' of level 4. | 3 / 5 |
Actionability | Concrete, executable commands throughout: 'moai harness status --project-root <project_root>', 'moai harness apply --execute --id <proposal-id> --project-root <project_root>', and a rollback invocation with a realistic timestamp. The payload schema is enumerated field-by-field. Minor gaps remain (the exact output format the skill should emit is described but not shown), so it sits at level 4 rather than fully copy-paste-ready level 5. | 4 / 5 |
Workflow Clarity | A clear five-step sequence with real checkpoints: status check first, execute path gated by an explicit --id (never inferred), the 5-Layer Safety Pipeline, and deletion confirmation on reject. It falls short of level 5 because there is no post-apply verification step (e.g., re-run 'moai harness status' to confirm the proposal cleared) for file-modifying and destructive paths. | 4 / 5 |
Progressive Disclosure | Well-organized single-file skill: Quick Reference up front, stepwise Implementation Guide, clearly signaled one-level-deep cross-references to contract files (askuser-protocol.md, harness foundation policy). No bundle files exist; the safety-architecture table and frozen-path list are compact enough inline but are candidate material for a reference file, keeping this at level 4 rather than the ideal split of level 5. | 4 / 5 |
Total | 15 / 20 Passed |