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.
The body is a strong, dense, actionable skill: a clear six-step workflow with explicit gates, a genuine retry/quarantine feedback loop, executable Playwright and CI code, and useful anti-pattern and limitations sections. The remaining weaknesses are minor: some repeated guidance across sections, a logically muddled quarantine snippet, and CI-example references (step id, ephemeral-deploy script) that don't resolve.
Suggestions
Fix the Step 4 quarantine snippet: the inner `steps.smoke.conclusion` check duplicates the outer `steps.smoke.outcome` condition and cannot distinguish flake from regression — instead capture the retry outcome (e.g., `--retries=2` result or last-attempt status) and branch on it, or drop the inner check.
Resolve the CI example's dangling references: add `id: deploy` to the ephemeral deploy step so `steps.deploy.outputs.url` works, and either ship `scripts/deploy-ephemeral.sh` in the bundle or mark it as a team-provided placeholder.
Deduplicate guidance repeated across sections (retry policy in Steps 2/4/Anti-patterns; read-only/pre-seeded data in Step 2/Anti-patterns/Limitations) — state each rule once and cross-reference, which would also make room to move the full CI workflow and product-type tables into one-level-deep reference files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by tables and executable code with almost no explanation of concepts Claude already knows, e.g., the Step 2 constraints table and the bare Playwright spec. Below the 5 anchor because of repetition: the 2-retry rule appears three times (Step 2 table "1 retry on transient failure" vs Step 4 "--retries=2" vs Anti-patterns "2 retries max"), and the read-only/pre-seeded-data guidance recurs in Step 2, Anti-patterns, and Limitations; above 3 because tightening would remove only these minor redundancies. | 4 / 5 |
Actionability | Quotes: the complete `e2e/smoke/checkout.smoke.spec.ts` Playwright spec, the full `.github/workflows/smoke-gate.yml` with `npx playwright test e2e/smoke/ --retries=2` — copy-paste ready for the common case. Below the 5 anchor due to minor gaps: the quarantine snippet's inner check of `steps.smoke.conclusion` duplicates the outer `steps.smoke.outcome` condition and never actually distinguishes flake from regression, and `BASE_URL: ${{ steps.deploy.outputs.url }}` reads an output from a deploy step that has no `id` (with `./scripts/deploy-ephemeral.sh` not shipped in the bundle); above 3 because the code is executable, not pseudocode. | 4 / 5 |
Workflow Clarity | Quotes: Steps 1-6 give a clear sequence (identify paths → implement fast → gate per stage → handle failure → curate → integrate CI) with explicit validation checkpoints ("A smoke failure halts the release", "timeout-minutes: 10 is the hard fail-fast") and a real feedback loop: retry twice, then either "block the deploy" for real regressions or "quarantine + investigate via flaky-test-quarantine". This matches the 5 anchor's explicit validation plus error-recovery loops; the destructive/batch cap does not apply since the skill creates tests rather than mutating data. | 5 / 5 |
Progressive Disclosure | No bundle files exist (no references/, scripts/, or assets/), so scoring is against the single-file structure: sections (Overview, When to use, Steps 1-6, Anti-patterns, Limitations, References) are well-organized and easy to navigate, and cross-skill pointers ("flaky-test-quarantine", "prod-canary-validator", "qa-web-e2e plugin") are clearly signaled. Below the 5 anchor because at ~210 lines the full CI workflow YAML and the product-type example tables are inlined where one-level-deep reference files would fit, and the body references a `scripts/deploy-ephemeral.sh` that does not exist in the bundle; above 3 because structure and navigation are genuinely good, not merely present. | 4 / 5 |
Total | 17 / 20 Passed |