Content
75%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, largely executable testing skill with strong vendor coverage (Cloudflare/Fastly/CloudFront), a clear canonical purge workflow, honest limitations, and a genuine one-level-deep reference split. It falls short of top marks on duplicated cache-status tables, an undefined helper in the multi-region example, sleep-based waits that contradict its own anti-patterns guidance, and CI-referenced scripts missing from the bundle.
Suggestions
Define or inline the `purge_url()` helper in the multi-region example, and replace the fixed `time.sleep(10)`/`time.sleep(30)` waits with the poll-until-`cf-cache-status: MISS` (timeout 30s) pattern the anti-patterns table itself prescribes.
Merge the 'Parsing results' table into the 'Cache-Status header verification' table — both explain the same cf-cache-status values.
Ship `scripts/write-test-fixture.sh` and `scripts/purge-test-paths.sh` in the bundle, or replace the CI run steps with the inline curl/pytest commands already shown in the Authoring section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with concrete curl/pytest/CI examples and assumes competence (no 'what is a CDN' padding), but the cf-cache-status values are explained twice — once in the 'Cache-Status header verification' table and again in 'Parsing results' — and the multi-region example partially repeats the canonical test. These minor redundancies keep it below the lean 'every token earns its place' anchor. | 4 / 5 |
Actionability | Vendor-specific curl commands, an executable canonical pytest, and CI YAML are copy-paste shaped, but the multi-region example calls an undefined helper `purge_url(API, ...)` and the CI steps invoke `./scripts/write-test-fixture.sh` and `./scripts/purge-test-paths.sh` which do not exist in the bundle. Concrete with minor gaps — not fully executable. | 4 / 5 |
Workflow Clarity | The canonical test lays out a clear numbered sequence (origin write → verify old edge state → purge → wait → assert fresh) with asserts at each step, and the anti-patterns table supplies recovery guidance. However the examples themselves use fixed `time.sleep(10)`/`time.sleep(30)` waits that the skill's own anti-patterns table says to replace with polling — a minor internal validation inconsistency. | 4 / 5 |
Progressive Disclosure | Browser-tier content is cleanly split into `references/browser-cache-control.md` and `references/playwright-cache-recipes.md`, both real files, clearly signaled and one level deep. The gap: the CI section references `./scripts/write-test-fixture.sh` and `./scripts/purge-test-paths.sh`, and no scripts/ directory exists in the bundle. | 4 / 5 |
Total | 16 / 20 Passed |