Content
78%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.
A tight, well-structured instruction-only subagent prompt with concrete matching rules and a copy-paste-ready output format. Minor gains are available from trimming the contextual framing and adding an explicit example exchange and clearer validation checkpoint.
Suggestions
Trim the opening framing paragraph about the ~30-entry cap to a single line; the 'Complete Skill Inventory' and 'Rules' sections already convey the purpose.
Add one short worked example (input keyword → expected ranked output) so the matching heuristic is unambiguous for edge cases.
Make the vagueness checkpoint explicit as a numbered workflow step (e.g., 'Step 0: If the task is too vague to extract 2-4 keywords, return the no-match fallback') rather than burying it in Rules.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Largely lean with no concept re-explanation and concrete rules, but the opening framing about the ~30-entry cap and 'look beyond that cap' is mild contextual padding that could be trimmed; sits just below the fully-lean top anchor. | 4 / 5 |
Actionability | Provides concrete, executable guidance — keyword count (2-4), explicit strong/weak match criteria, an exact output template, and a verbatim fallback string ('No installed skill matches: <task>') — though it lacks a worked input/output example to fully cover common cases. | 4 / 5 |
Workflow Clarity | A clear 3-step sequence (extract → match → return) for a read-only skill with no destructive ops, plus an implicit checkpoint ('If the task is vague, return what genuinely matched'); just short of a 5 because checkpoints are implicit rather than explicit. | 4 / 5 |
Progressive Disclosure | Under 50 lines with no external bundle files and clean, well-organized sections (Workflow, Rules) — the simple-skill exception applies and structure is fully navigable. | 5 / 5 |
Total | 17 / 20 Passed |