The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a user-specified TEST_COMMAND during the refactoring process to verify that code changes do not break existing functionality. This command is run by the worker agent at the start and after each logical change.\n- [DATA_EXPOSURE]: The hook script references/guard-idle-worker.sh reads from a transcript file path provided by the system via standard input to verify communication logs between the sub-agents before allowing a worker to go idle.\n- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes external source code from a user-provided TARGET_PATH. This code is ingested into the agent context for refactoring.\n
Ingestion points: Files and directories provided via the TARGET_PATH argument or interactive prompt.\n
Boundary markers: Absent. The worker and reviewer prompts do not define specific delimiters to separate the code being refactored from the agent instructions.\n
Capability inventory: The worker agent has file system write access within the target path and can execute the user-provided test command via the platform's terminal tools.\n
Sanitization: Absent. The skill does not perform any pre-processing or sanitization of the target files to detect or neutralize potential embedded instructions.

refactoring-team