Content
86%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.
This is a tight, single-purpose destructive skill executed well: concrete tool invocations with exact payloads, a scoped-purge decision workflow with confirmation checkpoints, and explicit schema error rules. The only refinements would be a brief error-recovery note for failed purge calls and deduplicating the irreversibility warnings.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — no concept explanations, just scopes, parameters, and rules — but the irreversibility warning recurs across the description, step 2, and the Important section, which could be trimmed. Not a 5 because of that minor redundancy. | 4 / 5 |
Actionability | Fully executable guidance: the exact tool ("mcp__context-mode__ctx_purge") is given with copy-ready parameter objects for every supported form ("{ confirm: true, sessionId: \"<id>\" }", "{ confirm: true, scope: \"project\" }", and the deprecated bare form), plus explicit schema rules for rejected combinations. | 5 / 5 |
Workflow Clarity | A clearly sequenced 4-step workflow with explicit validation for the destructive operation — user scope confirmation, a warning step, the mandatory "confirm: true" checkpoint, and result reporting — plus enumerated error conditions (rejected ambiguous params, throws without sessionId). Not a 5 because there is no guidance on recovering from a failed or errored purge call. | 4 / 5 |
Progressive Disclosure | The skill is under 50 lines, needs no external references, and no bundle files exist; per the simple-skill exception, its well-organized sections (Instructions, Schema rules, When to Use, Important) warrant a 5. | 5 / 5 |
Total | 18 / 20 Passed |