The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements logic in references/codex-imagegen.md to dynamically locate a TypeScript wrapper (main.ts) at runtime by traversing parent directories or checking environment variables. This script is then executed to facilitate image generation.\n- [COMMAND_EXECUTION]: The skill invokes system commands, specifically bun and npx -y bun, to run scripts located in sibling packages within the tool's directory structure. This is used as a fallback mechanism for image generation backends.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content from source.md and interpolates it into an image generation prompt template (references/base-prompt.md). While it includes a policy to strip credentials, the prompt construction lacks robust boundary markers to prevent the agent from following instructions embedded within the user-provided data.\n
Ingestion points: User-provided source content is stored in source.md and used to build the structured-content.md file.\n
Boundary markers: The references/base-prompt.md uses simple markdown headers and triple-dash separators but does not include explicit instructions for the AI to ignore instructions embedded in the interpolated content.\n
Capability inventory: The skill has access to multiple image generation tools (Codex imagegen, Cursor GenerateImage, and external CLI tools).\n
Sanitization: The skill contains a specific rule to remove credentials, API keys, and secrets from the output data before prompt generation.

baoyu-infographic