The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes annotation-source.json, which includes an aiPrompt field (e.g., "根据订单表格标注生成实现说明"). This field is explicitly intended to influence downstream AI processing during design reviews or delivery, creating a vector for indirect instructions if the data source is modified by an untrusted party.
Ingestion points: The annotation-source.json file and Markdown files in docs/*.md as described in SKILL.md.
Boundary markers: None identified in the instructions to separate data from instructions.
Capability inventory: React component rendering and local script execution (npm run, node).
Sanitization: No explicit sanitization or validation of the aiPrompt or markdown content is mentioned.
[COMMAND_EXECUTION]: The instructions require running local scripts using Node.js, specifically node scripts/check-app-ready.mjs /prototypes/<prototype-name>. This establishes a local command execution surface within the development environment.

prototype-annotation