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.
The body is an unusually actionable instruction-only skill: exact tool calls, exact strings, explicit edge cases, and a genuine verification checkpoint before completion. Its one weakness is redundancy in the Completion audit section, where the same evidence-standard rule is repeated across the bullet list and the closing paragraphs.
Suggestions
Consolidate the Completion audit: the two closing paragraphs repeat the bullet list's evidence standard, so merge them into one statement ('only mark complete when current evidence proves every requirement; otherwise keep working').
Trim duplicated phrases like 'weak, indirect, merely consistent with completion' which appear in both the audit bullets and the final paragraph.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean imperative guidance, but the Completion audit section restates the same 'do not mark complete without proof' idea several times ('Treat uncertain and indirect evidence as not achieved', 'Do not rely on intent, partial progress... as proof', 'If the evidence is incomplete, weak, indirect... keep working'), so it could be meaningfully tightened. This is more than the minor trimming of a 4 but well short of the padding of a 2. | 3 / 5 |
Actionability | The guidance is fully executable for an instruction-only skill: exact tool calls ('Call CreateGoal exactly once', 'call UpdateGoal with status "complete"', 'use TodoWrite'), exact user-facing strings ('Usage: /goal <objective>'), and concrete edge-case handling (empty objective, '30m'/'2h' prefixes, 'Every' routed to /loop). It is above a 4 because specific behaviors cover the common cases with no gaps. | 5 / 5 |
Workflow Clarity | The sequence (Parse -> Start -> Guidelines -> Completion audit) is clear, with explicit validation ('Before deciding that the goal is achieved, treat completion as unproven and verify it against the actual current state') and a feedback loop ('keep working instead of marking the goal complete'). Failure handling is present ('If creation fails, report that no goal was armed'), and the skill involves no destructive or batch operations that would cap the score. | 5 / 5 |
Progressive Disclosure | This is a self-contained single-purpose skill of about 50 lines with no bundle files and no content that belongs in separate references; its well-organized sections (Parse, Start, Guidelines, Completion audit) provide easy navigation, which satisfies the simple-skill standard for a top score. | 5 / 5 |
Total | 18 / 20 Passed |