blog-ingest

Pass

Audited by Gen Agent Trust Hub on Aug 29, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external, untrusted sources such as blogs and RSS feeds, which creates a surface for indirect prompt injection attacks. The skill documentation provides specific defensive procedures: (1) Ingestion points: Fetches article bodies from user-supplied URLs and feeds (SKILL.md); (2) Boundary markers: Mandates wrapping suspicious text in 'untrusted-quoted' blocks and setting 'untrusted_directives: true' in frontmatter; (3) Capability inventory: Uses 'fetch', 'gbrain search', 'put_page', and 'gbrain sync'; (4) Sanitization: Instructs the agent to treat all fetched text as data and to ignore embedded imperatives.
  • [COMMAND_EXECUTION]: The skill employs platform-native CLI tools ('gbrain search', 'gbrain sync') for knowledge base management, de-duplication, and synchronization. These commands are used according to standard operating procedures for the agent environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 29, 2026, 06:18 PM
Security Audit — agent-trust-hub — blog-ingest