uinaf/cross-harness-review
Run a read-only code review through the opposite AI coding harness and return uinaf review-style findings, evidence, unverified gaps, and a ship-it / needs-review / blocked verdict. Use for /review-with-claudex, cross-harness review, opposite-model review, Claude reviewing Codex work, or Codex reviewing Claude work.
92
92%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a strong description that clearly defines a specific niche (cross-AI-harness code review), lists concrete outputs (findings, evidence, gaps, verdict), and provides explicit trigger terms covering multiple natural phrasings. The negative boundary statement adds further clarity. The only notable flaw is what appears to be a typo ('uinaf' instead of likely 'unified'), which slightly reduces readability.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: 'run a read-only code review', 'return findings, evidence, unverified gaps', and 'a ship-it / needs-review / blocked verdict'. Also explicitly states what it is NOT ('Not an implementation or fix loop'). | 3 / 3 |
Completeness | Clearly answers both 'what' (run read-only code review, return review-style findings/evidence/gaps/verdict) and 'when' (explicit 'Use for' clause with multiple trigger scenarios). Also includes a negative boundary ('Not an implementation or fix loop'). | 3 / 3 |
Trigger Term Quality | Includes strong natural trigger terms: '/review-with-claudex', 'cross-harness review', 'opposite-model review', 'Claude reviewing Codex work', 'Codex reviewing Claude work'. These cover multiple natural phrasings a user might use. The term 'uinaf' appears to be a typo (likely 'unified') which slightly reduces clarity but doesn't undermine the trigger coverage. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive niche: cross-harness AI-to-AI code review with specific triggers like '/review-with-claudex' and 'opposite-model review'. The explicit exclusion of implementation/fix loops further distinguishes it from general code review or coding skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
79%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-crafted, actionable skill that efficiently teaches cross-harness code review with concrete CLI commands and a clear output contract. Its main weaknesses are the lack of explicit validation/error-recovery steps after delegation (e.g., what if the opposite harness returns malformed output?) and the inability to verify the referenced bundle file. The shared prompt template and dual output examples (success and blocked) are strong design choices.
Suggestions
Add a validation step after receiving the opposite harness's output (e.g., check that the response contains the required verdict field before reporting), which would improve workflow clarity for this multi-step delegation process.
Ensure the referenced file `references/headless-modes.md` exists in the bundle, or note its expected contents inline as a fallback.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is lean and efficient. It assumes Claude understands what a harness is, what diffs/PRs are, and how CLI tools work. No unnecessary explanations of basic concepts. Every section serves a clear purpose. | 3 / 3 |
Actionability | Provides fully executable CLI commands for both directions (Codex→Claude, Claude→Codex), a concrete shared prompt template, specific tool allowlists, and exact output format examples including the failure case. Copy-paste ready. | 3 / 3 |
Workflow Clarity | The workflow steps are clearly sequenced (define target → detect harness → preflight → delegate → report), but there's no explicit validation checkpoint after the review delegation step. The 'blocked' output example partially addresses failure, but there's no feedback loop for partial failures or malformed output from the opposite harness. | 2 / 3 |
Progressive Disclosure | References headless-modes.md for CLI flag details which is good progressive disclosure, but no bundle files were provided so we can't verify the reference exists. The skill itself is well-structured with clear sections, but the inline content for both harness directions could potentially be split into separate reference files for cleaner navigation. | 2 / 3 |
Total | 10 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
Reviewed
Table of Contents