The Agent Skills Directory

[SAFE]: The skill implements a secure method for handling API credentials by retrieving them from the environment (OPENROUTER_API_KEY) rather than hardcoding them in scripts or instructions.
[COMMAND_EXECUTION]: Orchestrates image generation by executing local TypeScript scripts (scripts/generate-image.ts) through the Bun runtime to manage network requests and file outputs.
[EXTERNAL_DOWNLOADS]: Connects to the OpenRouter API (openrouter.ai) to facilitate image generation based on user prompts. OpenRouter is recognized as a well-known service for AI model aggregation.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface through user answers in the 'Intake' stage, which are subsequently interpolated into a final image prompt.
Ingestion points: User-provided answers across 11 categories in references/stages/intake.md.
Boundary markers: Prompts are compiled into dense descriptive prose without explicit delimiter-based safety markers for the image generation model.
Capability inventory: Network access via fetch to OpenRouter and file writing capabilities to the Desktop via Bun.write in scripts/generate-image.ts.
Sanitization: Implements basic slugification for generated filenames to prevent path traversal or invalid character issues.

kernel-image