Content
86%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 concise, well-structured SSTI playbook with copy-paste detection probes and a clear escalation sequence. The weaknesses are concentrated in the escalation and validation sections, which rely on high-level gadget-chain direction and an undefined "validate_finding" tool rather than concrete payloads or an inline validation method.
Suggestions
Add at least one concrete escalation payload per major engine (e.g. a Jinja2 object-traversal chain for command execution) so the escalation path is as executable as the detection probes.
Replace or define "validate_finding" — either inline the validation method (how to confirm a positive execution signal and run a benign negative control) or bundle a script, since no validate_finding tool exists in the skill bundle.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean playbook: bullet-form sinks, per-engine probes, and a numbered escalation path with zero padding and no explanation of concepts Claude already knows. Every line carries actionable information, matching the "every token earns its place" anchor; there is nothing to trim or add. | 5 / 5 |
Actionability | Detection is fully executable — copy-paste probes per engine ("{{7*7}}", "${7*7}", "#set($x=7*7)$x") and concrete sink names ("render_template_string", "Template(...)", "twig->createTemplate"). Not a 5 because the escalation path stops at high-level directions ("Attempt command execution via framework-specific gadget chain") with no example payloads, and "validate_finding" references a tool that is not defined or bundled. | 4 / 5 |
Workflow Clarity | The escalation sequence (detect → enumerate → file read/env leak → command execution) is clearly ordered and a validation step exists ("positive execution signal and a benign negative control"). Not a 5 because there is no feedback loop for failed probes and "validate_finding" is an undefined reference; not a 3 because the sequence and a genuine validation checkpoint are both present. | 4 / 5 |
Progressive Disclosure | This is a short single-purpose skill (~20 lines) with no bundle files, so the simple-skill exception applies: the four well-organized sections (High-signal checks, Fast probes, Escalation path, Validation) are easy to navigate with no content that belongs in a separate file. All referenced items are inline — there are no dangling file references to verify. | 5 / 5 |
Total | 18 / 20 Passed |