The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes feature descriptions and requirements provided in the spawn prompt. This creates an indirect prompt injection surface where malicious instructions could be embedded in the input data to influence the agent's planning output. \n
Ingestion points: Feature descriptions and requirements are ingested from the initial spawn prompt. \n
Boundary markers: No explicit boundary markers or delimiters are defined to isolate untrusted input from the system instructions. \n
Capability inventory: The skill can execute subprocesses via uv run, write files to the plans/ directory, and send messages via the SendMessage tool. \n
Sanitization: No sanitization or validation of the input text is performed before it is used to generate plan files.\n- [COMMAND_EXECUTION]: The skill executes local Python scripts using uv run to perform self-validation of generated plans. \n
Execution method: Subprocess execution via uv run for scripts located in .claude/hooks/validators/. \n
Risk: While the scripts are local to the project's internal directory, this capability represents a vector for executing code if the validation scripts themselves were to be tampered with.

planner-workflow