The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through project-level configuration files.
Ingestion points: In SKILL.md, the agent is instructed to read 'Design Context' from local files such as .better-react-web-ui.md, .better-web-ui.md, .better-ui.md, and .impeccable.md located in the project root. The agent is directed to use this content as its authoritative design instruction.
Boundary markers: The instructions do not provide delimiters or clear 'ignore embedded instructions' warnings when processing these external project files.
Capability inventory: The skill empowers the agent to refactor and implement production code, which could be manipulated if an attacker places malicious instructions in the configuration files.
Sanitization: No sanitization, validation, or escaping of the content from these context files is required before use in the agent's prompt.

frontend-design