ado-gateway

Pass

Audited by Gen Agent Trust Hub on Oct 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the shell tool to execute a defined set of local bash scripts that manage the Azure DevOps interaction lifecycle. These scripts are strictly scoped to specific read and write operations and do not allow arbitrary command execution.
  • [EXTERNAL_DOWNLOADS]: The skill performs authorized network requests to official Microsoft Azure DevOps domains (dev.azure.com and vssps.dev.azure.com) using curl. These operations are required for its primary function of retrieving and updating development artifacts.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data from Azure DevOps pull request comments and work item fields. Ingestion points: Data enters via fetch scripts for PR comments and work items. Boundary markers: Output is delimited by a structured JSON handoff contract with explicit producer/consumer fields. Capability inventory: Access includes shell tool and network writes via curl POST requests. Sanitization: Implements automated HTML stripping and sensitive secret redaction in normalization scripts to mitigate injection risks.
  • [PRIVILEGE_ESCALATION]: A robust human-in-the-loop safety model is enforced for all mutation actions. Write scripts default to a dry-run mode that emits an action plan, and any network mutation requires an explicit --confirm flag from the user.
  • [DATA_EXPOSURE]: While the skill manages the AZURE_DEVOPS_PAT environment variable for authentication, it incorporates dedicated redaction logic in its write-guard and normalization scripts to prevent the accidental disclosure of tokens or bearer credentials in the agent's session output.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 3, 2026, 05:46 PM