Content
73%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-engineered diagnostic workflow with exceptional sequencing and validation logic, concrete commands, and copy-paste question scripts. Its weaknesses are repetition of the workflow-tier rationale across three sections and a handful of estimate fields (percentages, coverage) left without a stated method.
Suggestions
State the workflow-tier gap rules once (e.g. in step 3) and reference them from the opening blockquote and Follow-Up Actions instead of re-explaining the rationale each time.
Trim the persuasive asides (e.g. "the one thing it must never be... it would report Release for a project whose artifacts match its own Pre-Production row") to a single directive sentence.
Give a rough method for the report's [X%] completeness figures and the LOC/test-coverage estimates (e.g. bucket scales) so the template is fully computable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — it teaches project-specific rules rather than concepts Claude already knows — but the workflow-tier point is re-explained three times (the opening blockquote, step 3, Follow-Up Actions) and blockquotes like "a stage detector that cannot contradict its input is the one thing it must never be" are persuasive padding that could be trimmed. This goes beyond the "minor instances of over-explanation" of a 4. | 3 / 5 |
Actionability | Highly executable throughout: the exact invocation "bash .claude/scripts/artifact-check.sh" with constraints, per-directory checklists, copy-paste clarifying questions, a stage table, and a full report template. Not 5 because a few details remain unguided — the report's "[X%]" completeness figures have no computation method and "Estimate test coverage (rough heuristic)" / "Estimate lines of code (rough scale)" give no heuristic or scale. | 4 / 5 |
Workflow Clarity | Six clearly sequenced steps with explicit validation logic: a deterministic pass first, a forced compare of configured vs. observed stage with a named escalation path, a first-match-wins Stage Confidence ordering (FAIL > CONCERNS > NOT ASSESSED > PASS) with reasoning for each rank, and an approval gate before any write. This is a read-only diagnostic skill, so the destructive/batch validation cap does not apply. | 5 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ absent), and the body is well-sectioned with clearly signaled one-level-deep pointers to project docs (templates, automation-modes, code-root-resolution). Not 5 because at ~280 lines some self-contained material — the per-tier gap rules and the role-filtered recommendation lists — could live in reference files to keep SKILL.md a leaner overview. | 4 / 5 |
Total | 16 / 20 Passed |