The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external draft documents and user-provided original requirements. This creates a surface for indirect prompt injection if the source documents contain malicious instructions. However, the skill implements a strict cross-referencing logic (comparing IDs, roles, and status flows) which serves as a functional validation layer.
Ingestion points: Processes 'drafts' (초안) from development, planning, and screen agents, along with original text (원문) in SKILL.md.
Boundary markers: Uses specific labels like 설계제안: (Design Proposal) and 확인사항: (Verification Items) to separate implementation details from original requirements.
Capability inventory: Triggers external tools including documents:documents, pdf:pdf, figma:figma-generate-diagram, and product-ui for file generation and UI implementation.
Sanitization: The instructions focus on manual verification of rendering and logical consistency between requirements and output, though explicit text sanitization is not mentioned.
[COMMAND_EXECUTION]: The skill instructs the agent to invoke specific external skills (documents, pdf, figma, product-ui) to generate deliverables. These are standard platform tools for document processing and do not represent unauthorized command execution.

development-plan-review