Content
68%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, mostly lean instruction skill with concrete classification and output guidance, but its destructive refund workflow lacks explicit post-action verification and it gives no specific tool-invocation steps.
Suggestions
Add an explicit post-action verification feedback loop for destructive actions: after refunding or canceling, re-check subscription/invoice state to confirm the change took effect, and re-classify if it did not.
Include concrete tool-invocation guidance (specific Stripe API calls or portal actions) for each action in the preferred order so the operator knows exactly how to execute, not just what to do.
Consider moving the output-format template and the examples-of-good-recommendations into a separate reference file so SKILL.md stays a leaner overview and progressive disclosure can score higher.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Largely lean and directive with no padding explaining concepts Claude already knows; a few framing sentences ("The goal is to help the operator answer...", "Treat those as ECC or website follow-up items...") could be trimmed but add direction. | 4 / 5 |
Actionability | Concrete, actionable guidance via a classification table, preferred action order, and a copy-paste output template, but it stops short of specific tool/API invocations (e.g., which Stripe calls to make), leaving a minor execution gap. | 4 / 5 |
Workflow Clarity | Steps 1-5 are clearly sequenced with a pre-action classify checkpoint, but destructive refund operations lack an explicit post-action verify/feedback loop, which caps this at 3 per the destructive-operations guideline. | 3 / 5 |
Progressive Disclosure | Well-organized into clear sections (When to Use, Guardrails, Workflow, Output Format, Examples) with easy header navigation, but it is a single self-contained ~135-line file with no references, so it falls short of the split-content ideal. | 4 / 5 |
Total | 15 / 20 Passed |