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.
The body is an exemplary lean instruction skill: a concrete executable command, explicit success/failure semantics, safety boundaries around credentials and workspace writes, and a worked read-only example. The only gap is unspecified next steps when dsh exits one.
Suggestions
Add one line of failure-path guidance after the exit-code note, e.g. "On exit one, do not retry blindly: report DSH's stdout to the user and ask whether to adjust the prompt or retry."
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~25-line body is lean and efficient with zero padding: every sentence carries operational detail Claude would not know ("normally 10 minutes", "Pass the prompt as one quoted argument", "Treat stdout as final text, not JSON"), matching the anchor for assuming Claude's competence. | 5 / 5 |
Actionability | Guidance is fully executable: the exact command `dsh --profile headless "<prompt>"`, the availability check `command -v dsh`, explicit exit-code semantics (zero = success, one = failure), and a concrete read-only example with diff verification cover the common cases copy-paste ready. | 5 / 5 |
Workflow Clarity | The numbered sequence has explicit checkpoints (availability check, exit-zero confirmation, clean-diff inspection) and error paths for a missing binary or missing provider configuration. It stops short of a 5 because the exit-one failure path has no recovery guidance (retry, report, or adjust) after "exit one means the task failed". | 4 / 5 |
Progressive Disclosure | This is a simple under-50-line single-purpose skill needing no external references; the two well-organized sections (Run, Authentication And Permissions) keep navigation trivial, which the rubric's simple-skill guideline scores as 5. | 5 / 5 |
Total | 19 / 20 Passed |