The Agent Skills Directory

[SAFE]: The skill is a structured utility for performing adversarial reviews using multiple LLM models. It does not contain any malicious patterns or deceptive instructions.- [DATA_EXPOSURE]: No hardcoded credentials or sensitive file paths were detected. The skill interacts solely with content provided by the user for review purposes.- [REMOTE_CODE_EXECUTION]: The skill does not perform external downloads, package installations, or dynamic code execution. It uses platform-standard sub-agent spawning mechanisms for text generation.- [PROMPT_INJECTION]: The skill uses role-play instructions to guide sub-agents in providing critical feedback, which is the intended functionality. It includes clear delimiters for user-provided artifacts to reduce the risk of instruction confusion.- [COMMAND_EXECUTION]: No shell commands, privilege escalation attempts, or persistence mechanisms were found in the instructions or the accompanying evaluation files.

adversarial-review