The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically work item titles, descriptions, and comments retrieved from Azure DevOps.
Ingestion points: The skill performs GET requests to Azure DevOps API endpoints to retrieve existing work item data.
Boundary markers: The skill incorporates strong mitigations, including a mandatory "dry-run" (validateOnly=true) phase and a requirement for explicit user confirmation before any data is written or updated.
Capability inventory: The skill possesses network capabilities via curl and file-write access for temporary payload storage.
Sanitization: The instructions do not specify automated sanitization of retrieved HTML or Markdown content, relying on the user to review the changes during the confirmation step.
[DYNAMIC_EXECUTION]: The skill generates temporary files to handle complex data payloads in specific environments.
Evidence: For Windows PowerShell compatibility, the skill instructions state that the agent should write the JSON body to a temporary file (body.json) before sending it via curl.exe with the --data-binary flag.
[COMMAND_EXECUTION]: The skill relies on the execution of curl commands to interact with the Azure DevOps REST API.
Evidence: Multiple sections of SKILL.md provide shell command templates for interacting with dev.azure.com. These are legitimate and within the scope of the skill's primary purpose.

boards-work-item-write