The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow instructions in SKILL.md (Phases 2 and 3) direct the agent to load and execute Python scripts using the exec() function on content read from scripts/generate_plan.py and scripts/generate_todo.py. This method of dynamic loading is a high-risk pattern that can be used to execute arbitrary code if the filesystem is manipulated.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user input to generate task plans and markdown files without adequate safeguards.\n
Ingestion points: User-provided project goals or requests (processed in SKILL.md as user_intent).\n
Boundary markers: Absent; there are no delimiters or explicit instructions to the agent to disregard embedded commands within the user input.\n
Capability inventory: The skill writes to the filesystem (workspace/plan.json, workspace/todo.md) and executes Python logic within the agent's code execution context.\n
Sanitization: Absent; no validation, filtering, or escaping is performed on the user-provided string before it is interpolated into task objects and saved to files.

planning-task