implement-workitem

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from Azure DevOps work-item descriptions and PR feedback, creating a surface for potential injection attacks.\n
  • Ingestion points: Ingests data from work-item descriptions, comments, and PR feedback (SKILL.md, Workflow Step 1 and 2).\n
  • Boundary markers: Explicitly labels input as 'untrusted source material' and instructs the agent to disregard embedded commands (SKILL.md, Trust and authorization section).\n
  • Capability inventory: The agent can execute Git commands, interact with the Azure CLI, and perform code changes via plan-execution (SKILL.md, Handoffs section).\n
  • Sanitization: Requires verification of requirements against repository evidence and mandates multiple user approval checkpoints (SKILL.md, workflow checkpoints A through E).\n- [COMMAND_EXECUTION]: The workflow relies on executing shell commands for repository management and API interaction.\n
  • Evidence: Orchestrates git commands for worktree and branch management, uses the Azure DevOps CLI (az repos pr create) for PR creation, and runs a local script scripts/generate-handoff.py for payload normalization (SKILL.md and references/branch-and-publication.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 08:17 AM