ingest
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources such as web articles, social media threads, and meeting transcripts to update a persistent knowledge base.\n
- Ingestion points: The 'Media Workflows' section in SKILL.md identifies URLs, YouTube videos, PDFs, and social media posts as inputs.\n
- Boundary markers: No explicit delimiters or sanitization steps are defined to separate untrusted external content from agent instructions.\n
- Capability inventory: The skill uses put_page and gbrain sync (found in SKILL.md) to write to the file system and execute synchronization tasks.\n
- Sanitization: Not specified in the instructions.\n- [COMMAND_EXECUTION]: The skill uses the gbrain CLI tool to perform repository synchronization and file uploads.\n
- Evidence: SKILL.md includes commands such as gbrain sync --no-pull --no-embed and gbrain files upload-raw for managing data provenance.\n- [EXTERNAL_DOWNLOADS]: The skill fetches content from user-provided URLs and interacts with external services for transcripts and media storage.\n
- Evidence: The 'Media Workflows' and 'Raw Source Preservation' sections in SKILL.md describe fetching content from the web and interacting with supabase:// storage paths.
Audit Metadata