gemini-count-in-video

The skill claims to use the Google Gemini API for video object counting (which is consistent with the described purpose), but the package/registry metadata lists no required environment variables or credentials. The SKILL.md clearly expects a GEMINI_API_KEY (os.getenv('GEMINI_API_KEY')) and uses client.files.upload — the manifest should declare that credential. The absence of declared credentials is an incoherence between what the skill says it needs and what the registry requires.

The runtime instructions instruct the agent to upload video files (potentially surveillance footage) to Google Gemini, poll for processing state, and read local directories (e.g., /app/video) to batch-process files. Uploading sensitive video to an external API and reading local directories are within scope for a video-analysis skill, but the SKILL.md gives the agent broad filesystem access (os.listdir on a hardcoded path) and mandates prompt formats for model responses. The instructions do not limit which local paths to access or warn about sensitive data, and they rely on externally executing uploads — these are expected actions but should be made explicit in the manifest and docs.

This is an instruction-only skill with no install spec and no code files. That minimizes installation risk because nothing is downloaded or written at install time. No install-time downloads or archive extraction are present.

The SKILL.md uses GEMINI_API_KEY for genai.Client(api_key=os.getenv('GEMINI_API_KEY')), but the registry metadata lists no required environment variables or primary credential. That is disproportionate/inconsistent: a cloud API key is necessary for the described functionality and should be declared. Additionally, the instructions assume read access to a local directory (/app/video) and implicitly rely on local file availability — the manifest does not declare config paths or require consent for this access.

always is false and there are no install scripts or modifications to other skills or system-wide settings. The skill can be invoked autonomously (default), which is normal; this alone is not flagged. There is no evidence the skill requests persistent agent privileges beyond normal invocation.