The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process external specifications or requirements provided by users. This creates a potential surface where malicious instructions embedded in a specification could attempt to influence the generated implementation plan. However, the skill incorporates several safeguards:
Ingestion points: Technical specifications provided as input at runtime.
Boundary markers: The skill recommends including a path to the spec in the plan header and uses a structured template to separate plan logic from the source material.
Capability inventory: The skill is primarily focused on writing markdown files to the docs/superpowers/plans/ directory. It does not execute code directly, although it suggests tools for subsequent execution by other agents.
Sanitization: While no automated sanitization is described, the skill mandates a 'Self-Review' phase and a manual 'Execution Handoff' where the user must choose the execution method, providing opportunities for human oversight.
File System Operations: The skill is configured to save documentation to the docs/superpowers/ directory. This is a standard practice for development-oriented agents and is limited to project documentation paths.

writing-plans