The Agent Skills Directory

[COMMAND_EXECUTION]: The _spawn_agent function in scripts/persist.py invokes subprocess.run with shell=True. The command string is constructed using direct string replacement of the prompt file's content into a template (agent_cmd.replace('{}', prompt_text)). This allows any shell metacharacters in the prompt file to be executed by the shell.\n- [DYNAMIC_EXECUTION]: The skill generates and executes shell commands at runtime where a significant portion of the command string is sourced from a file (PROMPT.md) that is intended to be updated by an AI agent throughout the loop. This dynamic construction creates a high-risk execution path for data-driven payloads.\n- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect instructions being executed as commands:\n
Ingestion points: Prompt data is read from the file system by scripts/persist.py before being interpolated into a shell command.\n
Boundary markers: The skill does not use delimiters or robust quoting to prevent prompt content from escaping the intended command context.\n
Capability inventory: The script has the capability to run arbitrary shell commands with the user's permissions.\n
Sanitization: No escaping or validation is applied to the prompt content before it is processed in the shell environment.

persist