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 short-form skill: fully executable commands, explicit safety boundaries around authentication and permission escalation, and a validation step on the command output. The single gap is the absence of a stated recovery step for a failed run other than the missing-credentials case.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~27-line body is lean with zero padding: it never explains what OpenCode is or what a JSON event stream is, and every line (timeout guidance, availability check, command, parsing rule, credential boundary) earns its place. Matches the 'lean and efficient; assumes Claude's competence' anchor exactly. | 5 / 5 |
Actionability | Fully executable guidance: `command -v opencode` for availability, `opencode run "<prompt>" --format json`, a concrete 10-minute timeout, and an explicit rule to treat an error event or nonzero exit as failure, plus a worked example (inspect a failing test). Copy-paste ready with no gaps. | 5 / 5 |
Workflow Clarity | The Run section gives a clear 3-step sequence with checkpoints: set working directory and timeout, verify availability (with an explicit stop-and-ask on missing tool), run once, then validate output via error-event/nonzero-exit detection. Not a 5 because recovery beyond the credentials case is not specified — there is no validate-fix-retry loop describing what to do after a failed run. | 4 / 5 |
Progressive Disclosure | The skill is under 50 lines, single-purpose, and needs no external references; the two well-organized sections (Run; Authentication And Permissions) with the command inline satisfy the simple-skill exception. No bundle files exist, so nothing is misfiled or buried. | 5 / 5 |
Total | 19 / 20 Passed |