The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements file deletion using the rm -rf command. This behavior is the stated primary purpose of the skill and is constrained by safety logic.
Evidence: rm -rf "$RESOLVED" in scripts/safe-rm.sh is gated by an explicit --force flag and a protection check.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied file paths, creating a surface for potentially targeting sensitive files if input is not handled correctly.
Ingestion points: The TARGET argument passed to scripts/safe-rm.sh.
Boundary markers: Variable quoting is used throughout the script; instructions for the agent emphasize performing a dry-run first.
Capability inventory: rm -rf (file system modification).
Sanitization: The script uses path resolution (via cd, pwd, and sed) to normalize targets and checks them against a comprehensive blocklist (PROTECTED_PATHS) including system roots, home configurations (.ssh, .gnupg), and git repositories.
[SAFE]: No malicious patterns such as exfiltration, obfuscation, or persistence mechanisms were detected. The skill functions as a legitimate defensive utility for developers.

safe-rm