The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process user-supplied ideas and compile them into structured Markdown artifacts, such as implementation specs and build handoffs, which are intended for use by other AI agents.
Ingestion points: Untrusted content enters the workflow during the 'Capture' and 'Interview' stages (SKILL.md).
Boundary markers: The skill separates the workflow into discrete files (e.g., design-doc.md, implementation-spec.md), which provides structure, but lacks explicit formatting delimiters or 'ignore' instructions for the interpolated user content.
Capability inventory: The skill has permissions to write multiple Markdown files and load sub-skills for specialized tasks like UI design.
Sanitization: There are no documented processes for sanitizing, escaping, or validating the content provided by the user before it is inserted into the final handoff documents.
[PROMPT_INJECTION]: The skill documentation defines a 'User override phrase' (GREENLIGHT NEXT STAGE) which is explicitly designed to bypass the established clarifying interview process and progression rules. While the skill includes instructions to ensure this bypass does not affect safety filters or credential security, it represents a mechanism for overriding the agent's defined instructional logic.

idea-superpowers-suite