The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it transforms external specifications into structured plans containing executable commands.
Ingestion points: Requirements or specifications provided by the user to be processed by the writing-plans skill.
Boundary markers: The skill does not define explicit markers or instructions to treat input specs as untrusted or to isolate them from the plan generation logic.
Capability inventory: The generated plans include pytest commands, git operations, and Python code snippets meant for execution by sub-skills like superpowers:executing-plans.
Sanitization: No sanitization or validation logic is defined to prevent malicious commands from being included in the implementation plan if the source specification is compromised.

writing-plans