The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): No attempts to override agent behavior or bypass safety constraints were found. The skill is strictly instructional.\n- [DATA_EXFILTRATION] (SAFE): No unauthorized network operations or hardcoded credentials detected. A reference to /etc/passwd is included as a benign text-processing example for awk.\n- [COMMAND_EXECUTION] (SAFE): The skill provides code templates but does not execute commands itself. It proactively recommends against using eval with untrusted input to prevent vulnerabilities.\n- [REMOTE_CODE_EXECUTION] (SAFE): No external scripts are downloaded or executed. No unverifiable dependencies are referenced.

shell-scripting