Content
85%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 tight, well-structured debugging workflow with explicit sequencing, validation checkpoints, and an approval gate; its only weaknesses are minor — slight over-explanation in the rationalizations table and reliance on sibling skills/commands referenced by name without pointers or fallbacks.
Suggestions
Trim the Red Flags table's "Why It's Wrong" column (or merge it into "Do Instead") since the failure-mode reasoning is already evident to Claude, saving tokens without losing the guardrail.
Add a one-line fallback for when `dev-smoke`, `verify`, or `ai-devkit memory` are unavailable (e.g., a plain failing-then-passing test command), so the workflow stays executable outside the full suite.
Include a single worked example of a hypothesis entry showing the predicted evidence, disconfirming evidence, and exact test command format the "Hypothesize and Test" step requires.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and imperative, but the Red Flags table's "Why It's Wrong" column ("Assumptions skip evidence", "Stack traces show symptoms, not causes") explains failure modes Claude already understands, and the Task Tracing section hedges with "If task tracing is usable" — minor over-explanation that could be trimmed, fitting the score-4 anchor rather than the every-token-earns-its-place level. | 4 / 5 |
Actionability | Two copy-paste-ready commands (`npx ai-devkit@latest memory search/store ...`) and concrete option-comparison criteria (safer/simpler/more scalable) are strong, but sibling tools (`dev-smoke`, `verify`, `task`, `dev-retro`) are invoked by name with no pointer or fallback, and the required per-hypothesis "exact test command or check" is mandated without an example — minor gaps that keep it below fully executable. | 4 / 5 |
Workflow Clarity | A clearly sequenced Clarify → Reproduce → Hypothesize/Test → Plan workflow with explicit validation checkpoints: an approval gate ("Do not modify code until the user approves a selected fix plan"), "Confirm a pre-fix failing signal exists", and post-fix verification via the `verify` skill including regression checks, plus built-in feedback loops through disconfirming evidence. | 5 / 5 |
Progressive Disclosure | A self-contained ~62-line skill with clean sections (Hard Rule, Workflow, Validation, Task Tracing, Red Flags, Output Template); nothing is inlined that belongs in a separate file and no external references are needed, so well-organized sections alone earn the top score. | 5 / 5 |
Total | 18 / 20 Passed |