verify-llm-artifacts

Pass

Audited by Gen Agent Trust Hub on Jun 15, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses Python and shell commands (Git, test, ripgrep) to perform its verification tasks. These are executed locally and are part of the intended functionality for code analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from an external file (.beagle/llm-artifacts-review.json) which could theoretically contain malicious instructions if the source code being analyzed was designed to subvert the agent.
  • Ingestion points: Findings and descriptions are loaded from .beagle/llm-artifacts-review.json in step 1.
  • Boundary markers: The skill implements strict 'Hard gates' and an 'ID lock' mechanism in step 1b to ensure the agent only adjudicates findings explicitly listed in the source file.
  • Capability inventory: The skill uses python3 for data processing, git cat-file and test for file existence checks, and ripgrep for symbol searches across the repository.
  • Sanitization: The instructions include steps to echo the source content before adjudication and to validate the resulting JSON schema using Python, which helps prevent output corruption or hallucination.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 15, 2026, 06:03 PM
Security Audit — agent-trust-hub — verify-llm-artifacts