The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes shell scripts to perform environment cleanup and state management.
Uses rm -f to delete session-scoped and legacy state files within the .omc/state/ directory and user home directory.
Executes tmux kill-session and kill commands to stop background agent processes and workers.
Performs directory traversal via find and git commands to resolve state paths and locate active team configurations.
Includes an embedded Python one-liner (`python3
<<'PY' ...) to calculate future UTC timestamps for cancellation signals when local date` commands lack offset support.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and acting upon external configuration data and process lists.
Ingestion points: Reads team configuration files (config.json) and system process lists (ps aux) to determine targets for termination.
Boundary markers: None explicitly used; the skill parses data directly to find member names and process identifiers.
Capability inventory: Includes file system modification (rm), process termination (kill, tmux), and communication tools (SendMessage).
Sanitization: Uses sed to sanitize the project directory name for path construction, but relies on potentially untrusted data for other logic.

cancel