ado-gateway
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script
scripts/resolve-ado-user.pyuses__import__to load the standardosmodule for accessing environment variables. This use of dynamic loading targets a built-in module and does not involve untrusted or computed paths. - [DYNAMIC_EXECUTION]: The orchestration scripts
scripts/generate-handoff.pyandscripts/emit-handoff.py, as well as the test harnesstests/run-validation.py, utilizesubprocess.runto execute internal scripts within the skill's own directory. These executions are scoped to the skill's distributed assets and use the local Python interpreter. - [INDIRECT_PROMPT_INJECTION]: The skill fetches content from external Azure DevOps APIs, such as pull request comments and work item descriptions, which could contain malicious instructions. The skill implements several safeguards against this attack surface:
- Ingestion points: Data is fetched from Azure DevOps using authenticated GET requests in
scripts/fetch-work-item.pyandscripts/fetch-pr-comments.py. - Boundary markers: The skill emits a strictly validated JSON contract based on
assets/schemas/ado-openspec-handoff.schema.json, ensuring data remains structured. - Capability inventory: The skill possesses shell execution and Azure DevOps write capabilities (creating work items, PRs, and comments).
- Sanitization: The
scripts/normalize-work-item.pyscript strips HTML tags and normalizes text, whilescripts/_ado_lib.pyprovides centralized redaction of PATs and bearer tokens to prevent credential exposure in downstream tasks.
Audit Metadata