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 well-structured, token-efficient read-only subagent prompt: a clearly sequenced five-step workflow with named tools, an explicit stopping rule, and no padding. Its main gaps are the absence of any dead-end handling in the workflow and a vague final reporting step that gives no output format.
Suggestions
Add one fallback line to the workflow for dead ends, e.g. 'If grep returns nothing, broaden the pattern or list the matching directory with glob before giving up.'
Make the reporting step concrete by specifying an output shape (e.g. 'Report: 1-3 sentence answer, then the key files with paths and why each matters').
Merge the intro line and 'Your Mission' section (they both say fast, read-only exploration) to remove the remaining redundancy.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean, instruction-only body with no concept explanations Claude already knows. Minor redundancy: the intro line, "Your Mission", and "Your Restrictions" each restate the forked-subagent/read-only framing already established. Anchor 4 (efficient, minor trims possible); not 5. | 4 / 5 |
Actionability | Concrete tool-level instructions ("run `glob` with `**/*`", "use `grep` to locate key patterns", "use `file_reader` only on the files directly relevant"). Instruction-only absence of code is not penalized per scoring notes, but step 5 ("Report clearly — provide a concise, actionable summary") gives no output structure, leaving a minor gap. Anchor 4. | 4 / 5 |
Workflow Clarity | Five strictly ordered steps with a conditional ("Read README.md — if it exists") and an explicit stop condition ("Stop as soon as you have enough information"). The task is read-only, so the destructive/batch validation cap does not apply. Not 5 because there are no checkpoints or fallbacks for dead ends (e.g., what to do when grep returns nothing). | 4 / 5 |
Progressive Disclosure | No bundle files exist and the skill is under 50 lines with a single task; per the rubric's simple-skill guideline, well-organized sections alone warrant a 5. Sections (Mission, Restrictions, Workflow, Rules) are clearly headed and easy to navigate, and no inlined content belongs in a separate file. | 5 / 5 |
Total | 17 / 20 Passed |