ai-automation-workflows

Pass

Audited by Gen Agent Trust Hub on May 17, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The automation templates demonstrate ingesting external data (e.g., search results in content_pipeline.sh or file content in data_processing.sh) and interpolating it into LLM prompts. While standard for automation, this represents a surface for indirect prompt injection if the ingested content is from untrusted sources.
  • Ingestion points: content_pipeline.sh (research data), data_processing.sh (file content), conditional_workflow.sh (user input).
  • Boundary markers: None used in templates.
  • Capability inventory: Platform tools (belt), network requests (curl), file writes.
  • Sanitization: None demonstrated in examples.
  • [COMMAND_EXECUTION]: The skill provides scripts that use bash and Python subprocess to orchestrate platform tools. This behavior is consistent with the stated purpose of building automation workflows.
  • [EXTERNAL_DOWNLOADS]: Documents installation of the belt CLI and related skills from official vendor repositories on GitHub and NPM.
Audit Metadata
Risk Level
SAFE
Analyzed
May 17, 2026, 08:02 PM
Security Audit — agent-trust-hub — ai-automation-workflows