Content
70%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 process specification with explicit validation, pass accounting, and a repair feedback loop. Its weaknesses are the deferral of the core editing contract to external sibling paths that are not present in the bundle, making the skill not fully self-executable, and some redundancy plus inlined cross-skill contract material.
Suggestions
Inline or bundle the core editing contract (AI-writing pattern list, context exceptions, pass conditions, protected-region definitions) instead of deferring to ../avoid-ai-writing/SKILL.md, which is not part of this skill's bundle and leaves execution criteria missing.
Consolidate the three redundant failure-honesty statements (propagate write failures / do not claim edits from proposed patches / execution_evidence gating) into a single rule to tighten conciseness.
Move the cross-skill connection-contract and handoff-state machinery into a bundled reference file (e.g., references/handoff-contract.md) so the SKILL.md body stays a lean overview and the referenced paths actually resolve within the skill.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense and prescriptive with no explanations of concepts Claude already knows, but failure-honesty is stated redundantly ("propagate write failures instead of reporting success", "Do not claim a file was edited because a patch was merely proposed", and the execution_evidence gating rule), and the ~40-line connection-contract section could be tightened or externalized. | 4 / 5 |
Actionability | The process steps are concrete (pass counting via pass.index/pass.max, re-reading the changed region, reverting only affected spans), but the actual editing criteria — the pattern list, "context exceptions and pass conditions", and protected regions "defined by the canonical Skill" — are deferred to ../avoid-ai-writing/SKILL.md, which is not part of this skill's bundle, leaving key execution details missing. | 3 / 5 |
Workflow Clarity | The editing policy is a clearly sequenced 12-step process with explicit validation checkpoints (re-read after edit, VERIFY to preservation-verifier) and a complete feedback loop: FAIL from preservation-verifier triggers a bounded repair as the next editing pass, followed by re-verification and a stop-and-report rule if it fails again. | 5 / 5 |
Progressive Disclosure | Sections are well organized and references are one level deep and clearly signaled, but they point to sibling-skill paths outside the bundle that do not resolve (../avoid-ai-writing/SKILL.md, ../avoid-ai-writing-router/references/handoff-contract.md, skill-graph.json, agency-role-lenses.md), and the cross-skill connection-contract machinery is inlined rather than split out. | 3 / 5 |
Total | 15 / 20 Passed |