The Agent Skills Directory

[NO_CODE]: The skill consists entirely of documentation and instructional logic in Markdown format. There are no scripts (Python, Node.js, Shell), executables, or configurations that would lead to code execution.
[INDIRECT_PROMPT_INJECTION]: This skill functions as a judge, evaluating data produced by other agent actions. While it lacks specific boundary markers for sanitizing that data, it implements a strict evaluation logic and a mandatory checklist. This structure encourages the agent to verify coverage rather than being influenced by potentially malicious instructions embedded in the pentest results. Because the skill has no dangerous capabilities like network access or file manipulation, the risk profile remains safe.

judge-pentest