The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell logic in its runtime workflow to detect the presence of environment variables such as OPENAI_API_KEY and OPENAI_ORG_ID, as well as checking if a local binary defined by $OD_BIN is executable. This logic is used strictly for tier detection (Free, Pro, or Enterprise) to adjust its features accordingly.
[EXTERNAL_DOWNLOADS]: When configured for professional tiers, the skill makes outbound requests to api.openai.com to generate images. This is a well-known, trusted service and is used for the skill's primary intended purpose.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it reads project documentation to inform its design choices.
Ingestion points: It reads local project files including brand-guideline.md, prd.md, and package.json to extract brand identity details like colors and typography.
Boundary markers: It currently lacks explicit delimiters or specific instructions to ignore malicious commands that might be embedded within these documentation files.
Capability inventory: The skill's capabilities include reading project metadata, performing environment checks, and transmitting data to the OpenAI API.
Sanitization: There is no evidence of sanitization or filtering applied to the content extracted from project files before it is used to build prompts.

30-thiet-ke-master