The Agent Skills Directory

[SAFE]: The skill is a utility for security auditing. It implements a structured workflow for evaluating other agent skills for prompt injection, data exfiltration, and other malicious patterns.
[COMMAND_EXECUTION]: The skill uses read-only inspection commands (e.g., find, rg, sed, jq, file, git diff) and a dedicated security scanner (skillspector). These commands are necessary for the skill's primary function and are used in a controlled, non-destructive manner.
[INDIRECT_PROMPT_INJECTION]: While the skill processes untrusted input (the target skill being audited), the instructions include specific guardrails to mitigate risks, such as explicitly instructing the agent not to execute scripts from the target skill and treating it as untrusted data.

skill-inspector