Content
100%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.
An exemplary skill body: lean, fully executable, with a clearly sequenced workflow containing explicit validation and stopping criteria, and well-signaled one-level-deep references that match the actual bundle. No material weaknesses found.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line carries operational fact ("Collection does not require a binary", "-p 1 disables xdist", wrapper one-file-per-invocation behavior) with zero concept explanation or padding. Not 4 — there is nothing to trim. | 5 / 5 |
Actionability | Five copy-paste-ready wrapper invocations cover the common cases (single file, serial reproduction, parallel selection, collection-only, full suite), plus concrete `uv run --group dev ruff check/format` commands. Not 4 — the examples are complete and executable as written. | 5 / 5 |
Workflow Clarity | The workflow is clearly sequenced: run affected tests via the wrapper, rebuild via build-run if the binary is stale, fix regressions, lint changed Python files, check collection when imports/markers change, exercise parallel mode for fixture/port changes, and stop once checks pass. Validation steps (ruff, collection, "Once relevant checks pass") and a fix-regressions feedback loop are explicit; this is not a batch/destructive skill, so no cap applies. | 5 / 5 |
Progressive Disclosure | The body is a lean overview with two one-level-deep references, each clearly signaled by its trigger condition ("When adding, reviewing, or restructuring tests/helpers" → authoring.md; "For hangs, flaky tests, or parallel-only failures" → troubleshooting.md). Both files exist in references/, match their stated purpose, and contain no nested references; the only external link is a sibling skill (build-run), which is a boundary pointer, not a nested detail chain. | 5 / 5 |
Total | 20 / 20 Passed |