Content
92%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 instruction-only skill body: a crisp six-phase workflow with explicit validation and recovery loops, concrete defaults and path/subagent specifics throughout, and no token waste. The only weakness is a couple of implicit execution details (default candidate count, arena list configuration).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean, imperative prose with no padding and no explanation of concepts Claude already knows; the few rationale sentences ("The rationale is mandatory. Without it, the parent cannot tell…") exist to secure compliance, so every token earns its place. | 5 / 5 |
Actionability | Highly concrete for an instruction skill — default runners ("defaults opus, sonnet, haiku"), path template "/tmp/arena-<slug>/candidate-<n>/", "isolation: 'worktree'", "subagent_type: 'pstack:readonly-agent'", and a concrete-vs-vague rubric example — but minor gaps remain (no default N, and the "configured arena list" mechanism is unspecified). | 4 / 5 |
Workflow Clarity | Six clearly sequenced phases (Frame → Fan out → Cross-judge → Pick → Graft → Verify) introduced via a todolist, with explicit validation in Phase F and feedback loops for error recovery ("either Phase A was wrong (re-frame and re-run) or… go back to Phase E"), so the batch-operation validation cap does not apply. | 5 / 5 |
Progressive Disclosure | No bundle files exist and the body appropriately keeps all content inline in well-organized, clearly headed sections with no nested or buried references, which is the right structure for a skill of this size. | 5 / 5 |
Total | 19 / 20 Passed |