Skillv0.1.7

ClawScan security

PDF OCR Using Gemini LLM · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

BenignMar 7, 2026, 1:28 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's code, install requirements, and runtime instructions match its stated purpose (OCR via Google Gemini) and request only the single expected credential (GOOGLE_API_KEY).
Guidance: This skill appears to be what it says: it splits PDFs into single-page files and uploads them to Google Gemini for OCR, and it requires only GOOGLE_API_KEY. Before installing, consider: (1) privacy — full page images are sent to Google, so do not use with highly sensitive documents unless acceptable; (2) cost and quotas — large PDFs mean many uploads and API usage billed against your API key; (3) secure the GOOGLE_API_KEY (don’t paste it into logs or share it); (4) review and pin package versions if you want reproducible installs; (5) test on non-sensitive sample PDFs first to confirm behavior. If you need guarantees about retention or want OCR to run locally, consider a local OCR solution instead.

Purpose & Capability: okName/description, required env (GOOGLE_API_KEY), listed Python packages (google-genai, pymupdf), CLI entry point, and code all align with a PDF OCR tool that uploads pages to Google's Gemini API.
Instruction Scope: noteThe SKILL.md and code explicitly split PDFs into single-page files and upload full page files to Google's API for OCR. This behaviour is documented in the README and implemented in gemini_client.py (files.upload + models.generate_content). There are no apparent instructions or code that read unrelated files, other env vars, or send data to unknown endpoints, but note that entire page images are transmitted to Google (privacy/cost implication).
Install Mechanism: okDependencies are standard Python packages (google-genai, pymupdf, pydantic, pydantic-settings) and a requirements.txt is included. No downloads from custom URLs or extracts from arbitrary hosts are present.
Credentials: okOnly GOOGLE_API_KEY is required and declared as the primary credential. That single key is appropriate and required for the Google Gemini client used by the skill. No unrelated secrets or config paths are requested.
Persistence & Privilege: okThe skill is not always-enabled, does not modify other skills, and only writes temporary files under the system temp directory (cleans up after processing). It does not request elevated system persistence.