The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it processes untrusted user text and image references. Ingestion points: request structures for prompts and image URLs in SKILL.md. Boundary markers: Not explicitly present in prompt templates. Capability inventory: The skill provides code examples using urllib for network calls and os for file operations. Sanitization: Instructions explicitly mandate EXIF metadata stripping, magic byte verification, and human review of outputs.
[SAFE]: The skill demonstrates exceptional security awareness. It follows best practices for secret management by using environment variables and provides robust implementations for artifact acquisition that include HTTPS enforcement, redirect handling, and validation that host IP addresses are global to prevent SSRF attacks.

bytedance-seedream