Content
86%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 lean, well-structured body with executable commands, a clear five-step flow, and an explicit output format template. The main gaps are the missing error-recovery loop for failed validation and the absence of any concrete test-authoring example, though the latter is mitigated by the design's dynamic discovery of project testing conventions.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every section (Usage, Source of Truth, Execution Flow, Output Format) carries purely operational content — commands, a discovery step, an output template — with no explanation of concepts Claude already knows and nothing to trim, matching the lean 5-anchor. | 5 / 5 |
Actionability | Provides fully executable commands ('git diff --name-only', 'python3 ./.trellis/scripts/get_context.py --mode packages', 'pnpm lint', 'pnpm typecheck', 'pnpm test'), but the core test-authoring step is directed only as 'Add/update tests using existing project test patterns' with no example — a defensible gap given dynamic spec discovery, yet still short of copy-paste-ready 5. | 4 / 5 |
Workflow Clarity | Five clearly sequenced steps with an explicit validation checkpoint (step 4 runs lint/typecheck/test), but there is no error-recovery loop instructing Claude to fix failures and re-run validation, which the 5-anchor requires. | 4 / 5 |
Progressive Disclosure | A short, well-sectioned overview with no bundle files; the only external pointer (project '.trellis/spec/' unit-test files) is one level deep and clearly signaled with spec file names, qualifying for the simple-skill exception to score 5. | 5 / 5 |
Total | 18 / 20 Passed |