The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to invoke a local Python script located at .codex/skills/scripts/generate_image_smart.py using the uv command-line tool. This execution is used to render diagrams and illustrations based on the provided inputs.
[EXTERNAL_DOWNLOADS]: The skill utilizes the Pollinations.ai FLUX engine, a well-known service, to generate thematic concept illustrations. This involves fetching generated visual content from an external service during the execution process.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data via the content_spec input, which is interpolated into a shell command template, creating a vulnerability surface for indirect prompt injection.
Ingestion points: The content_spec field defined in the Agent Contract within SKILL.md receives instructions for diagrams or AI image prompts.
Boundary markers: There are no visible delimiters or specific safety instructions to the agent to disregard potentially malicious commands embedded within the content_spec input.
Capability inventory: The skill possesses the capability to execute subprocesses through uv run python to generate visual assets.
Sanitization: No explicit sanitization, validation, or escaping logic is described for the content_spec data before it is passed to the command-line interface of the generation script.

academic-figure-designer