ai-content-pipeline

Pass

Audited by Gen Agent Trust Hub on Jul 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous Bash examples for orchestrating AI workflows using the belt CLI. These commands are appropriately scoped within the allowed-tools frontmatter restriction.\n- [EXTERNAL_DOWNLOADS]: The documentation references official installation scripts and additional skills hosted on the inference-sh GitHub repository. These are standard resources for the platform's ecosystem.\n- [PROMPT_INJECTION]: The skill describes an orchestration pattern that naturally handles untrusted data between AI steps, representing a standard indirect prompt injection surface. This is the intended purpose of the workflow and is managed by platform-level guardrails.\n
  • Ingestion points: Outputs from intermediate steps (e.g., script.json, summary.json) are used as inputs for subsequent model calls.\n
  • Boundary markers: The examples focus on logic flow and do not include specific delimiters for data boundaries.\n
  • Capability inventory: The skill enables execution of inference apps via the belt tool.\n
  • Sanitization: No explicit sanitization is shown in the orchestration examples, as validation is typically handled by the individual inference applications.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 23, 2026, 06:04 PM
Security Audit — agent-trust-hub — ai-content-pipeline