Content
77%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 tightly written, governance-style instruction skill with strong sequencing, explicit validation checkpoints, and feedback loops across all five processes. Its weaknesses are the dangling reference to a `templates/TESTS.example.md` that is not shipped in the bundle (hurting both actionability and progressive disclosure) and some repeated assertions that could be trimmed.
Suggestions
Ship `templates/TESTS.example.md` inside the skill bundle (e.g. under `references/`) or inline a minimal worked example of a TEST-ID registry entry, so the output format Claude is told to adapt is actually visible.
Clarify where `templates/TESTS.example.md` and `.governance/docs-map.json` live (skill bundle vs. target project) — currently the reader cannot tell whether these paths exist or must be created.
Trim the repeated insufficiency statements: "test counts / file existence / green CI are not sufficient behavioral evidence" appears in Anti-Patterns, §5, and the closing paragraph; stating it once in §5 is enough.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a dense, declarative instruction set with no explanations of concepts Claude already knows, but a few statements are repeated: "Do not mirror every test function" appears in both Anti-Patterns and Goal, and the "test counts / green CI are not sufficient" point is made three times (Anti-Patterns, §5 checklist intro, and the closing line). This matches the 4 anchor's "minor instances of over-explanation that could be trimmed" rather than the 5 anchor's every-token-earns-its-place. | 4 / 5 |
Actionability | Guidance is concrete and executable as instructions: stable ID format (`TEST-ORDER-001`), controlled vocabularies for states, levels, and purposes, event-driven rescan triggers, a six-item confirmation checklist, and specific evidence commands per contract layer. It falls short of the 5 anchor because the artifact format itself is never shown — no worked example of a registry entry in SKILL.md, and the referenced `templates/TESTS.example.md` is not present in the skill bundle — leaving Claude without a copy-paste-ready shape for the output. | 4 / 5 |
Workflow Clarity | Each numbered process carries explicit validation checkpoints and feedback loops: §5's "Before marking a TEST-ID covered, confirm:" checklist (file exists, assertions verify behavior, exit code 0), "If the criteria are ambiguous, return to requirement clarification or ask the owner", leaving the TEST-ID `unverified` when a required gate is unmet, and the contract-first change loop (change contract → regenerate consumer assets → rerun provider and integration tests). Checkpoints are explicit and repeated, exceeding the 4 anchor's "most checkpoints present". | 5 / 5 |
Progressive Disclosure | The skill has no bundle files (no references/, scripts/, or assets/), and the body twice points to `templates/TESTS.example.md` ("The existing test registry and templates/TESTS.example.md"; "adapt templates/TESTS.example.md") with no such file in the bundle and no statement of where it lives — a dangling, unclearly signaled reference. That matches the 3 anchor ("references present but not clearly signaled"); it avoids a 2 only because the inline content itself is well-sectioned and appropriately placed rather than misfiled bulk material. | 3 / 5 |
Total | 16 / 20 Passed |