The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run standard development commands such as pnpm dev:status-page, pnpm lint, and pnpm typecheck. These commands are consistent with the skill's primary purpose of theme development, linting, and local testing within a specific repository structure.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of theme exports pasted by users to scaffold TypeScript configuration files.
Ingestion points: User-provided theme configuration strings (YAML or plain text) are pasted directly into the agent's context (referenced in SKILL.md and evals/evals.json).
Boundary markers: The skill provides a clear template in assets/theme-template.ts and instructions to clarify intent, but lacks specific instructions to escape or sanitize user-pasted values beyond following the TypeScript structure.
Capability inventory: The agent has the capability to write and modify files in the packages/theme-store/src/ directory and execute shell commands via pnpm.
Sanitization: The instructions advise the agent to preserve color values exactly as provided by the user, which presents a surface for potential injection if a malicious payload is designed to break out of the TypeScript object literal during scaffolding.

openstatus-theme