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, destructive-operation skill: lean prose, an exact executable command, a forced user-confirmation gate with post-run verification, and clear handling of the shared-memory and local-only variants. The only improvement would be brief guidance on what to do if the command fails or reports an error.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~20-line body contains no padding and no explanations of concepts Claude already knows; every sentence carries non-obvious information (command, flag semantics, confirmation rule, scoping of shared project memory). This matches the lean, every-token-earns-its-place anchor. | 5 / 5 |
Actionability | The single bash command is copy-paste ready and executable, and the body covers the common cases: remote delete with --yes, the --include-project-memory variant, and the local-only variant via omitting --remote. Matches the fully-executable top anchor. | 5 / 5 |
Workflow Clarity | The destructive operation has explicit validation: a confirmation gate before running ("After the user confirms"; "Never pass --yes before the user has confirmed") and a post-run verification step ("Report what the command output says was deleted"), so the destructive-op cap does not apply. It falls one short of the top anchor only because there is no error-recovery guidance if the command fails. | 4 / 5 |
Progressive Disclosure | The skill is under 50 lines with no content that belongs in separate files and no external references needed; the flow (scope, confirm, run, variants, report) is well organized, so the simple-skill exception for a 5 applies. | 5 / 5 |
Total | 19 / 20 Passed |