Content
73%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 well-structured, highly actionable workflow: concrete commands, explicit validation loops, and clear stopping conditions make it easy to follow autonomously. Its main weakness is redundancy — the same test/typecheck/lint cycle is stated three times across overlapping sections — which inflates token cost without adding information.
Suggestions
Collapse "Your Task", "Autonomous Workflow", and "Commands" into one canonical workflow: a single numbered sequence with the exact commands inline, removing the repeated test/fix/typecheck/lint instructions.
Merge "Before Coding" into the task sequence so planning steps appear once instead of duplicating steps 3-5.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body doesn't explain known concepts, but the same instructions recur across sections — run tests/fix failures, typecheck, and lint each appear in "Your Task", "Autonomous Workflow", and "Commands", and "Before Coding" repeats the planning steps — so it could be meaningfully tightened rather than just trimmed. | 3 / 5 |
Actionability | Concrete, executable guidance throughout: exact commands (`bun run test`, `bun run typecheck`, `bun run lint:fix`), project-specific conventions (`#src/*` import alias, `*.test.ts` co-location, `loadFixture()`), and a directory-structure example. Minor gaps: no example of how to locate or choose the spec file, and the only code block is a directory tree. | 4 / 5 |
Workflow Clarity | A numbered 9-step task sequence is reinforced by an explicit implement→test→fix→repeat loop with validation gates (tests pass, typecheck passes, lint passes) and clearly separated stopping conditions for both success and asking for help — matching the feedback-loop-and-checklist anchor. | 5 / 5 |
Progressive Disclosure | With no bundle files, all content appropriately lives inline under well-labeled sections, and the external references it does cite (ARCHITECTURE.md, CODE_STYLE.md) are clearly signaled. It misses a 5 because the body exceeds 50 lines and the duplicated sections could be consolidated. | 4 / 5 |
Total | 16 / 20 Passed |