The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements a robust interception mechanism using check-dangerous.sh to evaluate bash commands against a predefined set of security rules before they are executed by the agent.
[CREDENTIALS_UNSAFE]: Actively protects sensitive files by blocking read and write access to SSH keys, environment files (.env), and credential JSON files through logic in check-write.sh and specific rules in rules.json (e.g., G-030, G-035).
[REMOTE_CODE_EXECUTION]: Includes specific block patterns in rules.json (G-023, G-024, G-025) designed to detect and prevent the execution of remote scripts via piping curl or wget output to a shell.
[DATA_EXFILTRATION]: Features a secret redaction library (redact-secrets.sh) that sanitizes command strings before they are persisted to an audit log, preventing the leakage of inline secrets.
[PRIVILEGE_ESCALATION]: Blocks the use of sudo with destructive utilities and restricts modifications to critical system directories such as /etc/, /var/, and /usr/lib/.
[COMMAND_EXECUTION]: Employs recursive command analysis for command substitutions and eval bodies to ensure that nested commands do not bypass safety checks.

guard