data-scraper-agent
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The skill facilitates the collection of data from external websites and its transmission to external services including the Google Gemini API (via
generativelanguage.googleapis.com) and the Notion API. It manages sensitive environment variables such asGEMINI_API_KEYandNOTION_TOKEN. These operations are transparently documented as the primary function of the tool. - [COMMAND_EXECUTION]: The skill provides scripts and GitHub Actions workflows that execute shell commands to manage dependencies, run the orchestrator (
python -m scraper.main), and update the repository's state via Git. These commands are standard for the development and automation workflows described. - [INDIRECT_PROMPT_INJECTION]: The skill's architecture involves processing untrusted data from the web and passing it to an LLM, which represents an indirect injection surface.
- Ingestion points: Untrusted content is fetched from arbitrary URLs in
scraper/sources/source_name.py. - Boundary markers: The prompt construction in
ai/pipeline.pyutilizes basic Markdown headers to separate scraped items from instructions, which provides minimal isolation. - Capability inventory: The agent has the capability to write to external databases (Notion) and modify the source repository via GitHub Actions.
- Sanitization: The provided templates do not include explicit sanitization or filtering of the scraped text before inclusion in the LLM prompt.
Audit Metadata