The Agent Skills Directory

[SAFE]: The skill operates entirely through instructional prompts and markdown templates. It does not utilize any external tools, perform network requests, or access sensitive system files.
[PROMPT_INJECTION]: While the skill processes untrusted input in the form of a PRD.md file (Indirect Prompt Injection surface), its capabilities are strictly limited to documentation generation.
Ingestion points: Reads PRD.md as specified in SKILL.md and references/design-playbook.md.
Boundary markers: The instructions lack explicit XML tags or delimiters for the PRD content, but emphasize using 'only what the PRD contains'.
Capability inventory: Limited to reading local files and writing markdown documentation. No subprocess, network, or evaluation tools are used.
Sanitization: None detected, but the risk is negligible as the output is static documentation for human review.

system-design