The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines an extensive workflow that ingests user-provided task descriptions and project source code. This data is processed through stages that include technical analysis and automated execution, creating a surface for indirect prompt injection.
Ingestion points: User triggers and problem descriptions (Stage 1), and project files read during analysis (Stage 2).
Boundary markers: The skill enforces the use of specific output templates (defined in reference.md) for each stage to maintain structural integrity and delineate agent conclusions.
Capability inventory: Stage 6 (Execute) and Stage 7 (Verify) utilize powerful tools including file editing (Edit/Write) and shell command execution (Bash).
Sanitization: The skill relies on its multi-stage logic and confirmation gates rather than explicit string sanitization for external content.
[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions for installing missing dependencies from the author's repository using npx skills add FuDesign2008/open-skills. This is a vendor-owned resource used for legitimate setup purposes. It also references Figma (figma.com) for design-related verification tasks.

solve-workflow