data-pipeline
Pass
Audited by Gen Agent Trust Hub on Mar 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides various CLI commands for the user to execute pipeline stages and an orchestrator script (
./run_v2.sh). - [EXTERNAL_DOWNLOADS]: The pipeline performs network operations to fetch data from aggregators (Charity Navigator, ProPublica, etc.) and arbitrary charity websites via the
web_collector.pycomponent. - [PROMPT_INJECTION]: There is a vulnerability to indirect prompt injection due to the ingestion of external data.
- Ingestion points: Data is ingested from third-party websites by
web_collector.pyin thesrc/collectors/directory. - Boundary markers: No explicit use of delimiters or 'ignore embedded instructions' warnings are documented for the narrative generation prompts.
- Capability inventory: The skill uses an LLM to generate narratives, writes data to a Supabase database, and exports results to local JSON files.
- Sanitization: No security-focused sanitization of the retrieved web content is documented before it is passed to the LLM.
Audit Metadata