The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary shell commands through the task argument and the optional --check parameter. These commands are executed by the loop-agent using the Bash tool.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user-defined tasks and objectives that are subsequently passed to a background agent (loop-agent). This background agent is granted extensive capabilities, including file modification and shell access.
Ingestion points: The task and --until (objective) arguments provided via the /loop command in SKILL.md.
Boundary markers: No explicit delimiters or boundary markers are defined when interpolating user inputs into the background agent's context.
Capability inventory: The loop-agent.md defines access to Bash, Write, Edit, Read, Grep, and Glob tools, enabling significant file system and system-level operations.
Sanitization: The instructions do not specify any validation, escaping, or filtering for the user-provided task or objective strings before execution.
[DYNAMIC_EXECUTION]: The initialization process invokes a local Python script scripts/cv_init_store.py. This script utilizes subprocess.run to execute git commands (rev-parse) to determine repository paths and calculate a project slug for state management. It also performs file system operations including directory creation in the user's home directory and symlink creation within the repository.

loop