The Agent Skills Directory

[SAFE]: No security issues were detected. The skill performs standard project onboarding tasks.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a common vulnerability surface by ingesting external data (project documents and source code) into the agent context.
Ingestion points: Reads files in design/gdd/, .claude/docs/, and src/ (SKILL.md, Phase 1).
Boundary markers: Absent; the agent reads raw file content without explicit delimiters or instructions to ignore embedded commands.
Capability inventory: Uses Write to manage production/review-mode.txt and AskUserQuestion for user interaction.
Sanitization: None; the skill assumes processed files contain valid documentation or code.
Assessment: This is a standard and expected pattern for a project discovery skill. The lack of network capabilities or high-privilege operations mitigates the risk of this attack surface.

start