Content
65%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill is highly actionable with extensive executable examples, but it is verbose and redundant, lacks a true sequenced workflow with validation checkpoints, and is a monolithic wall of text with no progressive disclosure into separate reference files.
Suggestions
Collapse the duplicated Quick Reference payload/encoding tables into the inline code blocks they repeat, and trim explanations of basic concepts (what HTML injection is, HTML-vs-XSS differences) that Claude already knows.
Restructure the Phases into an ordered testing workflow with explicit validation steps — e.g., after batch fuzzing in Phase 9, require manually verifying each flagged reflection before reporting it as a finding.
Move the payload catalogs (Phases 3, 4, 8) and the remediation guide (Phase 10) into separate reference files (e.g., references/payloads.md, references/remediation.md) and keep SKILL.md as a concise overview that links to them.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is substantive but padded: it restates basics Claude already knows ("HTML injection occurs when user input is reflected... without proper sanitization") and the Quick Reference tables duplicate payloads already shown in Phases 3 and 8. | 2 / 3 |
Actionability | Provides copy-paste-ready executable guidance throughout: curl probes, a runnable Python fuzzing script, concrete Burp/ZAP steps, and PHP/Python/JavaScript remediation code. | 3 / 3 |
Workflow Clarity | The ten "Phases" read as a topical catalog rather than a sequenced operational workflow, and Phase 9 (automated batch fuzzing) lacks the validation/feedback checkpoints that the rubric requires for batch operations, capping this at 2. | 2 / 3 |
Progressive Disclosure | No bundle files exist and the entire ~490-line skill is a single monolithic SKILL.md; payload catalogs and remediation guidance that should live in one-level-deep reference files are inlined, though sections are reasonably organized. | 2 / 3 |
Total | 9 / 12 Passed |