Content
57%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 dense, technically credible attack checklist with genuinely actionable commands and payloads, undermined by systematic redundancy (the same vulnerability list appears four times), no verification checkpoints in the testing workflow, and a monolithic single-file structure that should offload mobile, remediation, and alternatives detail to reference files.
Suggestions
Consolidate the redundant vulnerability enumerations: the Quick Reference bullets, the Vulnerability Map ASCII tree, and the Vulnerabilities section all restate the same class list — keep one map and let the detailed sections carry the specifics.
Add validation checkpoints to the Manual Testing Steps (e.g., after each technique, what response signature indicates success — HTTP status deltas, signature-accepted vs rejected behaviors — and when to re-test with a modified payload).
Move the Mobile App JWT Storage, Remediation Recommendations, and Alternatives & Modern Mitigations sections into reference files (e.g., references/mobile-extraction.md, references/remediation.md) and link them from a short overview, keeping SKILL.md itself a lean checklist.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The same vulnerability classes are repeated across four sections (Quick Reference bullets, the Vulnerability Map tree, the Vulnerabilities section, and Manual Testing Steps), and the Mechanisms section explains basics like 'JWTs (RFC 7519) consist of three Base64URL-encoded parts' that Claude already knows. Mostly efficient but could be tightened significantly by consolidating. | 3 / 5 |
Actionability | Mostly executable guidance with copy-paste-ready commands ('python3 jwt_tool.py <token> -M all', 'adb shell "run-as com.target.app cat ..."', concrete kid/jwk/jku header payloads). Minor gaps: step 1 is pseudocode ('base64url_decode(header).payload.signature') and the timing-attack script needs target adaptation. | 4 / 5 |
Workflow Clarity | 'Follow steps in order' plus the ordered Hunt and Manual Testing Steps give a clear sequence, but validation checkpoints are entirely absent — there is no guidance on how to confirm a technique succeeded (response differences, error signatures, re-testing). | 3 / 5 |
Progressive Disclosure | No bundle files exist and the entire ~280-line body is inlined in SKILL.md. Section headers are clear, but content that belongs in separate reference files (mobile storage extraction detail, remediation recommendations, alternatives/modern mitigations) is inlined, and the file far exceeds the 50-line simple-skill case. | 3 / 5 |
Total | 13 / 20 Passed |