Content
71%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-structured, evidence-first review skill: clear sequencing, concrete inspection targets, a strict output contract, and appropriately split reference files. The main defects are unresolved required links to the shared policy framework and an unreferenced validation script, plus prose that could be tightened in the workflow sections.
Suggestions
Resolve or remove the dead required links: the nine shared/base/*.md policy files and shared/workflow-contract.md referenced in the Base Framework and Optional Workflow Integration sections do not exist in the bundle, so either ship them or inline only the policies actually needed.
Wire the existing scripts/validate_scenarios.py into the workflow (e.g., a final step: run `python scripts/validate_scenarios.py` to confirm the output contract and scenario coverage), turning the quality checklist into a mechanical feedback loop.
Break the long multi-clause sentences in workflow steps 3, 4, and 6 into short scannable bullets (mechanism, impact, trade-off, validation) to cut token load without losing the guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and directive, naming SOLID/DDD/CQRS without explaining them and reserving tokens for skill-specific judgment calls ("Optimize for independent change ... not abstraction count"). It is not a 5 because some multi-clause sentences in the workflow steps and the Base Framework policy-link block carry tokens that could be trimmed without losing guidance. | 4 / 5 |
Actionability | Concrete guidance throughout: specific inspection targets ("imports, constructors/DI wiring, ORM/framework types, database calls"), named tools (DepCruise/Madge/Nx/ArchUnit/jQAssistant, ripgrep), an exact output-section contract, required per-finding fields, and three worked examples. It falls short of anchor 5 because there are no runnable commands and the bundle's own scripts/validate_scenarios.py is never referenced as an executable step. | 4 / 5 |
Workflow Clarity | The six-step workflow is clearly sequenced with real checkpoints: Verified/Likely/Assumption/Unknown classification, the "Unscored" rule for scores lacking rubric and evidence, and a closing quality checklist. It does not reach anchor 5 because there is no validate-and-recover feedback loop and the checklist is not tied to any mechanical verification step. | 4 / 5 |
Progressive Disclosure | The three references/*.md files are real, one level deep, and well-signaled at the matching workflow steps and examples. However, the body's "Required packaged policies" block links nine shared/base/*.md files plus shared/workflow-contract.md that do not exist in the bundle (and the validation script expects an absent agents/openai.yaml), so required references fail to resolve — more than the "minor organization gaps" of anchor 4. | 3 / 5 |
Total | 15 / 20 Passed |