Content
72%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-sequenced skill body built almost entirely from executable ffuf/gobuster/curl commands that cover the technique comprehensively. The two real weaknesses are the missing validation of batch-enumeration results (no false-positive/soft-404 baseline or confirmation step, which caps workflow clarity) and complete failure to reference the existing scripts/agent.py and references/api-reference.md bundle files, leaving them orphaned.
Suggestions
Add a false-positive baseline step to the enumeration workflow (e.g., probe a random nonexistent path to calibrate soft-404 response size/status and use ffuf -fs auto or -fl to filter), plus an explicit 'manually confirm each finding before reporting' checkpoint.
Wire up the bundle: add a step or note pointing to `python scripts/agent.py --target <url> --admin-paths --session-cookie <token>` as the automated path, and link to references/api-reference.md for the full argument tables and function documentation.
Trim the Key Concepts table (Claude already knows these terms) and drop or fold the Feroxbuster/DirBuster rows from Tools & Systems, since neither tool is used anywhere in the workflow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The workflow itself is lean — each step is a one-line lead-in followed by executable commands — but the "Key Concepts" table (Forced Browsing, Directory Enumeration, Path Normalization, etc.) and "Tools & Systems" table restate knowledge Claude already has, and the Tools table lists Feroxbuster and DirBuster that never appear in the workflow. These are minor, trimmable instances of over-explanation rather than the 'some unnecessary explanation' of score 3, since the bulk of the body is compact, purposeful command blocks. | 4 / 5 |
Actionability | Nearly every step is copy-paste-ready: ffuf commands with real wordlist paths, match codes, output files, thread/rate flags; gobuster invocations; shell loops for actuator endpoints, HTTP methods, and sensitive files; concrete curl comparisons for authenticated vs unauthenticated access. Coverage spans directories, files, extensions, admin paths, API endpoints, actuator, method override, path normalization, double encoding, and backups, matching the 'fully executable, covers the common cases' anchor. | 5 / 5 |
Workflow Clarity | The six steps are clearly sequenced (enumerate → find admin/debug → test auth enforcement → method bypass → path bypass → backup discovery) and Step 3 includes an explicit verification heuristic ("If both return similar content, authentication is not enforced"). However, the batch enumeration operations have no result-validation guidance: there is no soft-404 / custom-error-page baseline check (a classic false-positive source for ffuf-based discovery) and no manual-confirmation checkpoint before reporting findings. Per the guideline capping batch-operation workflows without validation at 3, this sits at score 3 rather than 4, though it is above score 2 because the sequence and one checkpoint are genuinely present. | 3 / 5 |
Progressive Disclosure | The body is well-sectioned (When to Use, Prerequisites, numbered Workflow, Scenarios, Output Format) and stays a reasonable length, but it never references the bundle files that actually exist: references/api-reference.md and scripts/agent.py are completely orphaned — no link, no mention, no navigation — despite automating and documenting exactly what the body teaches. Content that could live in the reference file (command catalogs, argument tables) is inlined. This matches 'some structure but could be better organized' with references effectively unsignaled; not score 2 because the in-body structure itself is solid and the body is not an unstructured wall of text. | 3 / 5 |
Total | 15 / 20 Passed |