Content
93%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 patterns reference: dense with executable code and decision rules, no filler, and a properly layered bundle where deep material (sweepers, ephemeral testing, full check catalog) is delegated to real one-level-deep reference files. The only notable gap is the absence of how to execute the tests it teaches the reader to write (TF_ACC / go test invocation).
Suggestions
Add a short 'Running tests' note covering TF_ACC=1 and the `go test -run TestAcc...` invocation so the write-test workflow closes the loop with an execution and failure-feedback step.
Consider mentioning `resource.UnitTest` / `-tfacc-module-test-dir` or at least one sentence on where PreCheck failures surface, to make the lifecycle's validation checkpoints actionable from the CLI side as well.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and directive — field tables instead of prose, terse rules like "Use `resource.ParallelTest` by default. Use `resource.Test` only when tests share state" and "Do not mix `Check` (legacy) and `ConfigStateChecks` in the same step" — and it never explains concepts Claude already knows (no "what is Terraform" padding). Every section earns its tokens; fits the 'lean and efficient' anchor rather than score 4's 'minor instances of over-explanation'. | 5 / 5 |
Actionability | Fully executable, copy-paste-ready Go for every common case: a complete test function, provider factory, config helpers with numbered format verbs, all five scenario patterns, custom exists/disappears StateCheck implementations, the shared state-lookup utility, a CheckDestroy TestCheckFunc, and PreCheck. Minor example-specific placeholders (example_widget) are appropriate for a patterns skill; not score 4 since no key detail is missing. | 5 / 5 |
Workflow Clarity | The test lifecycle is explicitly sequenced ("plan → apply → refresh → final plan", then destroy and CheckDestroy, with the failure condition stated), and scenario patterns give ordered step compositions including the two-commit regression workflow. It falls short of 5 because there is no guidance on actually running the tests (e.g., TF_ACC=1, `go test` invocation) or a run → read failure → fix loop, leaving a checkpoint gap. | 4 / 5 |
Progressive Disclosure | A clearly signaled "**References** (load when needed)" block lists three real, one-level-deep files with one-line descriptions (`references/checks.md`, `references/sweepers.md`, `references/ephemeral.md` — all verified present), and topics whose depth lives in those files (sweepers, ephemeral/echoprovider, full knownvalue catalog) are correctly kept out of the body. Matches the 'clear overview with well-signaled one-level-deep references' anchor; not 4 because navigation and content split are clean. | 5 / 5 |
Total | 19 / 20 Passed |