classify-work-items

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the Azure CLI command az account get-access-token to obtain authentication tokens for interacting with Azure DevOps resources.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting external finding data. Ingestion points: Untrusted data is read from findings.json. Boundary markers: The skill uses structured JSON schemas defined in references/data-contracts.md to scope data processing. Capability inventory: The skill performs command execution for authentication, makes network requests via Invoke-RestMethod, and writes to backlog_input.json. Sanitization: The skill mitigates injection risks by explicitly instructing the agent to escape HTML characters (&, <, >) in the current, expected, recommendation, and notes fields before including them in the final work item description.
  • [EXTERNAL_DOWNLOADS]: The skill performs authenticated REST API requests to official Azure DevOps endpoints (dev.azure.com) to discover project process templates and metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 02:26 AM
Security Audit — agent-trust-hub — classify-work-items