Content
81%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-crafted, opinionated instruction skill: concrete stamps, labels, and commands, a mandatory human validation gate before publishing, and a tight anti-patterns section. The remaining room is trimming rhetorical reinforcement, adding one worked ticket example, and considering whether the tier/domain rubrics belong in a reference file to slim the main body.
Suggestions
Add a short worked example of a finished ticket body (title, Blocked by / Tier / Domain lines, ready-for-agent label) so the publish format is copy-paste ready.
Trim repeated or rhetorical justification — e.g. the second 'never a model name' passage in the domain rubric section and the 'agent asked to size itself' aside in rule 9.
The body runs ~64 lines; moving the tier and domain rubrics into a references/ file (linked from the rules) would keep SKILL.md a lean overview and earn the top progressive-disclosure anchor.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence (no explanations of what a PRD or ticket is), but includes rhetorical expansions that could be trimmed — e.g. 'an agent asked to size itself has every incentive to answer "the strongest one"' — and restates the never-name-a-model rule twice ('Still never a model name... the rot rule 9 avoids'). This fits 'efficient; minor instances of over-explanation that could be trimmed' rather than the every-token-earns-its-place level. | 4 / 5 |
Actionability | Highly concrete guidance: exact stamps ('Tier: <planner|implementer|mechanical|reviewer>', 'Domain: <token>', 'Blocked by: #N', 'Part of #<prd>'), the ready-for-agent label mechanism, a first-hit-wins tier rubric, and a real command ('gh issue create'). The minor gap keeping it below 'fully executable... specific examples cover the common cases' is the absence of a worked example of a finished ticket body. | 4 / 5 |
Workflow Clarity | The procedure is a clear 5-step sequence with an explicit validation gate before the batch, hard-to-reverse publish step ('Quiz step (mandatory human gate)... Do not publish until they confirm'), a stop-and-surface rule for untrusted PRD content, and an error-recovery heuristic ('A decomposition that came out all one tier is a finding worth stating'). This matches the top anchor: explicit validation steps with feedback loops for the risky publish operation. | 5 / 5 |
Progressive Disclosure | A single well-organized file with clear sections (Rules, tier rubric, domain rubric, Trust boundary, Procedure, Anti-patterns) and repo-level pointers (docs/domain-glossary.md, scripts/agents.config.sh) rather than nested skill references. At ~64 lines it exceeds the under-50-line threshold where a single organized file would merit a 5, and the tier/domain rubrics could arguably live in a reference file, so it fits 'good structure; minor organization gaps'. | 4 / 5 |
Total | 17 / 20 Passed |