The Agent Skills Directory

[SAFE]: The skill is a design interview tool that assists users in planning AR projects. It is explicitly instructed not to write code or modify project files, and it does not access sensitive system data or perform unauthorized network operations.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the transfer of user-provided project descriptions to a downstream building agent (CLAD) via a design brief file. This architecture represents a potential surface for instruction injection from the user to the secondary agent.
Ingestion points: User chat responses provided during the AR design interview process (SKILL.md).
Boundary markers: The generated design brief (devpost/design-brief.md) uses markdown headers for structure but lacks explicit instruction-boundary warnings for the downstream agent.
Capability inventory: This skill writes local markdown files. The downstream consumer (CLAD) has build and testing capabilities within the Lens Studio environment.
Sanitization: The skill incorporates a human review process where the user must review the draft brief and provide explicit sign-off ("looks good") before the brief is finalized, serving as a manual validation step.

plan-lens