Content
53%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 well-structured, largely actionable manual testing playbook whose core weakness is its relationship to its own bundle: the shipped automation script and its API reference are never referenced from SKILL.md, so a reader would never know they exist. Secondary issues are explanatory padding of known concepts and absent validation checkpoints (scope confirmation, finding reproduction, post-test cleanup) around batch and data-modifying test operations.
Suggestions
Add a section wiring in the bundle, e.g., '## Automated scanning: For bulk checks run `python scripts/agent.py --base-url https://api.target.com --token <jwt> --low-priv-token <jwt> --graphql --output report.json`; see [references/api-reference.md](references/api-reference.md) for all functions and OWASP coverage.'
Insert validation checkpoints into the workflow: confirm written authorization and test accounts before Step 1, reproduce each finding with a second request before reporting, and restore/verify any data modified during mass-assignment or DELETE-method tests.
Trim the Key Concepts table and Tools & Systems entries of definitions Claude already knows (JWT, rate limiting, Burp Suite, Postman), keeping only skill-specific methodology terms like BOLA/BFLA test technique.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The workflow steps are dense and directive ('Change `alg` to `none` and remove the signature', 'hashcat -m 16500 jwt.txt wordlist.txt'), but the Key Concepts table and Tools & Systems section explain concepts Claude already knows (JWT as 'a self-contained token format used for API authentication', Rate Limiting, Burp Suite, Postman), adding ~20 lines of padding that could be trimmed. | 3 / 5 |
Actionability | Mostly executable guidance: copy-paste-ready GraphQL introspection query, mass-assignment JSON payload, NoSQL operator injection payload, and a hashcat command, plus concrete endpoint examples ('GET /api/users/123/orders -> GET /api/users/456/orders'). Minor gaps keep it below fully executable — e.g., 'Test all parameters... with SQL injection payloads' and 'Submit XML content with external entity declarations' give no example payloads, and SSRF testing lists no concrete metadata addresses. | 4 / 5 |
Workflow Clarity | The five-step sequence (Discovery, Authentication, Authorization, Input Validation, Data Exposure) is clearly ordered, but there are no validation checkpoints around the batch and destructive-natured operations the workflow itself prescribes — 'Send 100+ rapid requests to authentication endpoints', 'DELETE /api/users/456', and mass-assignment writes that modify data. Per the rubric, missing validation/verification for batch/destructive operations caps this at 3. | 3 / 5 |
Progressive Disclosure | The bundle ships scripts/agent.py (a 228-line automated testing agent) and references/api-reference.md (its CLI and function documentation), yet the SKILL.md body contains zero references to either — no path, no link, no mention. The bundle is effectively buried and undiscoverable from the main file, which is the navigation failure this dimension measures; the body's own section structure is otherwise good. | 2 / 5 |
Total | 12 / 20 Passed |