The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill orchestrates the execution of local project code using Python's unittest or custom test wrappers. This involves running code contained within the project worktree to verify hypotheses.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from the local environment. Ingestion points: Project source files, test files, and JSON measurement data. Boundary markers: No explicit delimiters are defined to separate untrusted content from the agent's instructions. Capability inventory: The skill executes local code and shell commands via a launcher. Sanitization: There is no evidence of sanitization for the project-provided test outputs or files.
[COMMAND_EXECUTION]: The skill uses a platform-specific CLI launcher to perform operations such as starting loops, checking status, and accepting changes, which involves shell-level interaction with the host environment.

bounded-loop