Content
77%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 highly actionable, well-structured hunting skill with concrete payloads, a clear methodology, and strong validation gates. Main weaknesses are inline-heavy organization (no progressive disclosure to reference files) and some explanatory prose that could be trimmed for token efficiency.
Suggestions
Move the payload catalog, parser-ecosystem vulnerability matrix, and disclosed report citations into separate reference files under ./references/ and link to them from SKILL.md, keeping the main file a lean overview with one-level-deep navigation.
Trim the 'Common Root Causes' enumeration and the per-scenario 'Business impact' prose — Claude already understands why XXE is severe; retain only the payload/root-cause specifics that change hunting behavior.
Consolidate the duplicated classic file-read payload that appears in both 'Payload & Detection Patterns' and the bypass sections to reduce token cost and maintenance drift.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient reference material (payloads, grep patterns, parser matrix), but the 'Common Root Causes' list and the 'Real Impact Examples' business-impact prose explain impact Claude already understands and could be tightened. | 3 / 5 |
Actionability | Fully executable: copy-paste XML payloads, curl commands for the JSON→XML content-type swap, language-specific grep patterns, and a concrete 10-step methodology with specific endpoints and URI-wrapper schemes — common cases are covered comprehensively. | 5 / 5 |
Workflow Clarity | Clear 10-step hunting sequence with explicit validation checkpoints (Pre-Severity Gate inline-entity probe, Gate 0 Validation) and feedback loops (no-reflection→blind OOB, OOB-blocked→error-based exfiltration), matching the top anchor. | 5 / 5 |
Progressive Disclosure | Well-organized with clear section headers, but ~400 lines are entirely inline with no bundle files in ./references/, ./scripts/, or ./assets/; the payload catalog, parser-ecosystem matrix, and disclosed report citations clearly belong in separate reference files. | 3 / 5 |
Total | 16 / 20 Passed |