Content
63%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 skill's core verification-gate content is strong: a memorable iron law, a concrete 5-step gate, evidence-requirement tables, and runnable examples. Its weaknesses are redundancy (the Red Flags table duplicates the Rationalization Table), a jargon-heavy opening section that doesn't connect to the rest, and a dangling reference to a host-adapter file that does not exist in the bundle.
Suggestions
Merge the 'Rationalization Table' and 'Red Flags' sections into one — they repeat the same excuses nearly verbatim and could cut ~15 lines.
Fix or remove the reference to 'skills/blocks/codex-host-adapter.md' (and the sibling-skill list): either ship those files in references/ or drop the cross-references, since they currently point at files not in the bundle.
Rewrite or move the 'Portable feature evidence' opening section: its jargon ('policy provenance', 'open-marker count', 'task-contract digest') is not actionable and is not connected to the Gate workflow below it.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly operational rather than explanatory, but there is clear redundancy: the 'Red Flags — STOP and Verify' table restates the 'Rationalization Table' almost line-for-line ('I'm confident', small-change excuses, agent-reported success appear twice), and the host/multi-provider material ('Claude Octopus', 'orchestrate.sh', probe/grasp/tangle/ink) is niche context that could be tightened. This fits 'mostly efficient but includes some unnecessary explanation or could be tightened' — better than noticeably verbose (2), short of the efficient (4) anchor. | 3 / 5 |
Actionability | The Gate gives a concrete executable sequence (IDENTIFY/RUN/READ/VERIFY/STATE), the Evidence table maps specific claims to specific required outputs, there are runnable bash commands for synthesis verification, and the red-green regression example is fully concrete. It falls short of 5 because the opening 'Portable feature evidence' section is abstract jargon ('policy provenance and open-marker count through the existing boundary adapter') with no executable counterpart, and most verification commands appear as examples rather than copy-paste-ready prescriptions. | 4 / 5 |
Workflow Clarity | The core workflow is a clearly sequenced 5-step gate with explicit validation checkpoints and error-recovery guidance (rationalization handling, red-green cycle with revert/re-run), plus a concrete when-to-apply checklist. It misses 5 because the 'Portable feature evidence' opening section doesn't integrate into that sequence and reads as disconnected requirements, leaving a minor coherence gap in an otherwise strong flow. | 4 / 5 |
Progressive Disclosure | The body references 'skills/blocks/codex-host-adapter.md' and names sibling skills (flow-develop.md, skill-tdd.md, etc.), but none of these files exist in the bundle — the reference is dangling, so navigation leads nowhere. At ~135 lines the file is also above the under-50-line simple-skill case, and host-specific adaptation notes are inlined in the body rather than split out. Some structure exists (well-labeled sections), matching the 3 anchor: 'some structure but could be better organized; references present but not clearly signaled.' | 3 / 5 |
Total | 14 / 20 Passed |