fix-ado-comment-markdown
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
az(Azure CLI) command via thescripts/ado_comments.pyscript to interact with Azure DevOps REST APIs. This is a core functionality for posting, patching, and fetching comments. The execution is performed usingsubprocess.runwith a list of arguments and does not useshell=True, which is a secure implementation that prevents shell injection vulnerabilities. - [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from Azure DevOps work item discussions. When the
pullcommand is used, existing HTML comments are converted to Markdown drafts for the agent to review and edit. This creates an attack surface where maliciously crafted comments could contain natural language instructions aimed at the agent. - Ingestion points: Comment text is fetched from Azure DevOps via REST API calls in
scripts/ado_comments.py. - Boundary markers: While the skill instructions encourage manual review and comparison with original text, it lacks explicit prompt delimiters to strictly isolate the ingested content from agent instructions.
- Capability inventory: The skill has file system write access for drafts/manifests and network access via the
az restcommand to update Azure DevOps work items. - Sanitization: The conversion process includes HTML unescaping and basic tag removal, which provides formatting sanitization but does not filter for semantic instructions within the text.
Audit Metadata