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.
The body is a lean, well-organized reference with mostly executable detection code, a useful severity table, and concrete defender mitigations. Its main gaps are the absence of an explicit step sequence with validation checkpoints (verify the differential reproduces before rating severity, verify PoC across sessions) and a placeholder pseudocode snippet in the XS-Search section.
Suggestions
Add a short numbered PoC workflow with a validation checkpoint, e.g., '1. Pick a technique from section 1; 2. Build the attacker page; 3. Verify the differential reproduces across authenticated vs. unauthenticated sessions; 4. Only then rate severity per section 5.'
Replace the placeholder comment in the 1.7 XS-Search loop with executable code (e.g., a concrete timeFetch call over guessable_set with a threshold comparison).
Move the 7-category taxonomy detail into a references/ file (e.g., techniques.md), keeping SKILL.md as a concise overview that links to it, to strengthen progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient — tight one-to-three-line technique descriptions and small runnable snippets with no padded sections or library-comparison fluff. Not 5 because the opening conceptual framing ('XS-Leaks abuse browser primitives that leak information ACROSS origins... Smaller than full cross-origin read') and the section-2 prose ('Decepticon agents should consult this...') are mild over-explanation that could be trimmed. | 4 / 5 |
Actionability | Mostly executable guidance: the image-timing snippet, the frame-count detection code, timeFetch, and the exact defender headers (COOP/COEP/CORP, X-Frame-Options, Vary: Cookie) are concrete and runnable. Not 5 because section 1.7 is placeholder pseudocode ('// measure timing or frame-count differential') and sections 1.3, 1.5, and 1.6 give only prose directions with no code; not 3 because the majority of guidance is copy-paste ready. | 4 / 5 |
Workflow Clarity | Section order loosely implies a process (categories -> detection -> PoC framing -> severity -> defender), but there is no explicit sequence and no validation checkpoints — e.g., nothing instructs verifying the leak differential across authenticated vs. unauthenticated sessions before rating severity, and section 4's 'Visual demonstration of leaked info' is only an implicit checkpoint. The batch ID-enumeration workflow (1.7) likewise has no verification step, consistent with the cap at 3. | 3 / 5 |
Progressive Disclosure | No bundle files exist (no references/, scripts/, or assets/), and the ~110-line body is well-sectioned and self-contained, with a clearly signaled external reference (xsleaks.dev plus what it provides) and explicit cross-references to related skills. Not 5 because the body exceeds 50 lines with the full 7-category taxonomy inlined — it could be split into a references file with SKILL.md as overview; not 3 because structure is clear and nothing is buried. | 4 / 5 |
Total | 15 / 20 Passed |