The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data, including text prompts and image files, which are transmitted to the Ideogram API for generation and editing. This introduces a surface for indirect prompt injection where malicious instructions could be embedded in the input data.
Ingestion points: User-provided text_prompt, json_prompt, image, mask, and style_reference_images fields defined in SKILL.md.
Boundary markers: The documentation explicitly references Ideogram's safety filters (is_image_safe flag) and usage policy compliance as primary guardrails.
Capability inventory: The skill uses requests for network operations, Pillow for image validation/processing, and sqlite3 for local persistence of webhook deliveries in its provided Python examples.
Sanitization: Recommends redacting IDEOGRAM_API_KEY from logs, using secret managers, and performing client-side validation of file sizes and dimensions before processing.
[SAFE]: The skill follows security best practices for credential management, advising the use of environment variables and secret managers for API keys rather than hardcoding. It also provides detailed implementation for Ed25519 signature verification of webhooks using official JWKS endpoints, which is a high-standard security pattern.

ideogram-image