The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (scripts/generate-icon-reference.sh and scripts/verify-icon.sh) to automate icon generation and verification. These scripts are internal to the skill and perform specific tasks related to the icon modernization workflow.
[EXTERNAL_DOWNLOADS]: The generate-icon-reference.sh script makes HTTPS requests to OpenAI (api.openai.com) and Google (generativelanguage.googleapis.com) to generate image references. These are well-known services used for the skill's primary function. The script handles API keys through environment variables and implements timeouts and safety checks on the responses.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user-defined icon concepts which are then interpolated into prompts sent to external image generation APIs.
Ingestion points: User-supplied icon descriptions in SKILL.md workflows are passed as arguments to the generation script.
Boundary markers: The script wraps the concept in a structured prompt but lacks explicit "ignore instructions" delimiters for the interpolated content.
Capability inventory: The skill has the capability to execute shell scripts and perform network operations to specific well-known APIs.
Sanitization: The script uses jq to safely construct the JSON payload for the API requests, which prevents basic JSON structure manipulation, although it does not sanitize the natural language content for potential prompt injection against the remote image model.

typo3-icon14