The Agent Skills Directory

[CREDENTIALS_UNSAFE] (LOW): The scripts/check_setup.py script prints a preview of the GEMINI_API_KEY (first 20 and last 4 characters) to stdout. While not the full key, this practice leads to unnecessary credential exposure in execution logs.
[COMMAND_EXECUTION] (MEDIUM): scripts/check_setup.py implements dynamic code loading by adding ~/.claude/scripts to the Python search path and importing resolve_env. Loading executable logic from computed external paths is a security risk if the environment is not strictly controlled.
[PROMPT_INJECTION] (LOW): The skill is vulnerable to Indirect Prompt Injection through the processing of untrusted media.
Ingestion points: scripts/gemini_batch_process.py and scripts/document_converter.py ingest user-provided images, PDFs, audio, and video.
Boundary markers: No delimiters or instructions to ignore embedded prompts are used when passing content to the model.
Capability inventory: The skill makes remote API calls to Gemini models which can be influenced by malicious content in the media.
Sanitization: No content sanitization or validation of the media's visual/auditory instructions is performed.

ai-multimodal