Content
75%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 disciplined, operational instruction skill: it sequences the review workflow with verification checkpoints, gives exact commands and output contracts, and avoids padding almost entirely. Remaining improvements are structural — breaking up the dense Quality Gates paragraph and folding the two-axes pass into the main workflow — rather than substantive.
Suggestions
Split the single-paragraph Quality Gates section into a short list or steps so its conditions (no entries vs. configured, BLOCK vs. advisory) are scannable.
Integrate the two-axes sub-agent pass into the numbered Workflow (e.g. as a step conditional on diff size) so the sequencing is unambiguous.
Add one brief example of a well-formed finding (file, line, failure scenario, severity) to make the output contract concrete.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — no explanations of what diffs or reviews are, and lines like "A plausible-sounding finding that does not reproduce is noise" earn their tokens — but a few justificatory clauses and slight redundancies ("Advisory by default — review informs, it does not gate") could be trimmed, matching the efficient-with-minor-excess anchor rather than the every-token-earns-its-place anchor. | 4 / 5 |
Actionability | Guidance is concrete and executable: the exact command `omc quality-gates list --stage omc-review --json`, verdict strings `VERDICT: PASS`/`VERDICT: BLOCK`, the record shape `{ gate, mode, verdict }`, named standards-discovery targets (CLAUDE.md/AGENTS.md, docs/standards/, CONTRIBUTING.md, CODING_STANDARDS.md), and a defined output format ("file, line, and concrete failure scenario"); the minor gap is the absence of a worked example of a finding, which keeps it below the fully copy-paste-ready anchor. | 4 / 5 |
Workflow Clarity | The five numbered steps include an explicit validation checkpoint ("Verify each candidate finding before reporting it") and the Quality Gates section adds real feedback loops (BLOCK → REQUEST CHANGES, unresolved → gate_unavailable, reusing the gate lane on re-review), but the two-axes sub-agent pass sits outside the numbered Workflow list, leaving mild sequencing ambiguity — clear sequence with minor gaps rather than a fully integrated checklist. | 4 / 5 |
Progressive Disclosure | The skill is self-contained with no bundle files and clean, well-labeled sections (Goal, Workflow, What to check, Two axes, Rules, Output, Quality Gates) and no nested references; however, the body runs ~53 lines and the Quality Gates section is one dense unbroken paragraph, a minor organization gap that keeps it just below the well-organized-simple-skill top anchor. | 4 / 5 |
Total | 16 / 20 Passed |