The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to review user-provided code files via the /ui-skills <file> command, which creates a surface for indirect prompt injection. Malicious instructions embedded within the source code being reviewed could attempt to influence the agent's behavior. No explicit boundary markers or sanitization instructions are included in the skill definition to mitigate this risk.
Ingestion points: User-supplied source code files processed via the review command in SKILL.md.
Boundary markers: Absent.
Capability inventory: Command execution capabilities via the opencode environment as specified in eval.yaml.
Sanitization: Absent.
[SAFE]: The skill uses established, well-known libraries for UI development, including Tailwind CSS, Framer Motion (motion/react), and accessible primitives like Base UI and Radix. These are standard industry tools and do not represent a security risk.
[SAFE]: The evaluation scripts (graders/check.py) and reference solutions (solutions/reference-checkout-card) use standard system utilities and the Python standard library for static analysis and file generation without any network activity or suspicious command execution.

ui