The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the CLIP library from OpenAI's official GitHub repository. It also lists standard dependencies from the public PyPI registry, including torch, torchvision, pillow, ftfy, regex, and tqdm.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it processes external, untrusted data inputs.
Ingestion points: Both SKILL.md and references/applications.md contain code to load local image files using Image.open() and tokenize user-provided text queries via clip.tokenize().
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the code examples to mitigate against adversarial content in inputs.
Capability inventory: The skill uses machine learning libraries (torch, clip, transformers) to compute embeddings. It lacks high-risk capabilities such as arbitrary shell command execution, system-level file modifications, or non-whitelisted network communications.
Sanitization: No input validation or sanitization logic is provided in the example implementations for image or text data.

clip