Content
71%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, practical attack playbook with real commands, payloads, tools, a decision-driven PoC workflow, and a useful severity table. Its weaknesses are a missing when-to-use framing in the description side, minor conceptual padding, schematic XML examples, and a monolithic single-file layout with cross-references to paths that don't exist in the bundle.
Suggestions
Split detail-heavy material (full XSW1-8 mechanics, known exemplars/CVEs, severity calibration) into references/ files inside the bundle and replace the dangling `skills/_corpus/...` cross-references with links to those real files.
Add the missing execution glue: a command or snippet for re-encoding a modified SAMLResponse (base64 + deflate) and re-submitting it, making the PoC pattern fully end-to-end.
Trim the opening background paragraph and the explanatory prose in sections 5-8 down to one-line statements of the attack condition, letting the XML examples carry the explanation.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is largely lean — terse sections, XML payload snippets, and a severity table with no library surveys or padding. Minor over-explanation keeps it from a 5: the opening conceptual framing ("SAML is XML-based SSO used heavily in enterprise. Signatures protect assertions but XML's structural flexibility...") and explanatory prose in sections 5-8 restate things Claude already knows and could be trimmed to one line. | 4 / 5 |
Actionability | Concrete, executable guidance dominates: `echo "$SAMLResponse_base64" | base64 -d | xmllint --format -`, a copy-paste comment-injection payload, named tools with specific usage ("SAML Raider → 'XSW Test' → cycle through all 8 variants"), and a step-by-step PoC pattern. It falls short of 5 because the XSW1 XML uses schematic placeholders ("..."), XSW2-8 mechanics are deferred entirely to the tool, and no command shows how to re-encode/re-submit a crafted SAMLResponse. | 4 / 5 |
Workflow Clarity | Section 11's PoC pattern is a clear 7-step sequence with an explicit success check ("Any variant that returns logged-in-as-target = bug") and fallback loops ("If XSW fails, try comment injection... If still fails, try signature stripping"), so the destructive-operation cap does not apply. Not a 5 because the workflow lacks a validate→fix→retry checkpoint structure (e.g., how to verify the crafted message decodes/re-encodes correctly before submitting, or how to document the resulting session beyond a vague step 7). | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are absent), and the "Cross-references" section points to `skills/_corpus/payloads/SAML Injection/` paths outside this bundle that are not navigable files within it. The 145-line body is well-sectioned but monolithic — detailed XSW variants, severity calibration, exemplar CVEs, and remediation are all inline where a references/ file split would fit. This matches the anchor of some structure with references that are present but not clearly signaled/useful. | 3 / 5 |
Total | 15 / 20 Passed |