Content
90%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, highly actionable single-file skill that respects the token budget and gives concrete detection, exploitation, and remediation guidance for every major framework. Its only real gaps are the absence of an explicit failure-recovery loop in the detection workflow and minor opportunities to split wordlist/exemplar content into reference files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean throughout — framework patterns are one-to-three-line snippets, severity is a compact table, and the field names are a flat wordlist. No section explains concepts Claude already knows; every token earns its place. | 5 / 5 |
Actionability | Concrete HTTP request examples, a field-name wordlist, five framework-specific vulnerable code patterns, a copy-paste Python PoC covering register → login → assert is_admin → confirm admin, and remediation code make the guidance fully executable across the common cases. | 5 / 5 |
Workflow Clarity | Numbered sections (Detect → fields → frameworks → leak → tools → PoC → severity → defender) form a clear sequence with real validation checkpoints ('Re-fetch own profile. If any of the injected fields persists... → mass assignment'; the PoC's assert step). It falls short of anchor 5 because there is no explicit error-recovery loop for silently rejected injections — retry guidance is only implicit in the section 2 wordlist. | 4 / 5 |
Progressive Disclosure | A single-file skill (~145 lines) with no bundle directories and well-organized, clearly headed sections. Structure is good and content is mostly appropriately placed, but items like the field-name wordlist and known-exemplars section could live in one-level-deep reference files, and there are no reference files at all, so it does not fully match anchor 5. | 4 / 5 |
Total | 18 / 20 Passed |