The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the automated, recurring execution of arbitrary commands or prompts specified in the user arguments. While it implements safety controls (max iterations, circuit breaker, minimum interval), the underlying capability relies on the agent's powerful Bash and Agent tools to carry out the repeated tasks.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it parses and executes commands directly from the $ARGUMENTS variable. If these arguments are populated with untrusted data from external sources (such as a pull request description or file content), it could lead to the execution of unintended actions within the repetition loop.
Ingestion points: $ARGUMENTS parsed in SKILL.md and processed in the execution protocol.
Boundary markers: Absent; the skill does not wrap the user-provided command in protective delimiters or provide explicit warnings to ignore embedded instructions.
Capability inventory: Employs Bash, Write, and Agent tools across its execution protocol, allowing for diverse and powerful actions.
Sanitization: No validation, escaping, or filtering is performed on the command string before it is passed to the execution tool.

repeat