The Agent Skills Directory

[SAFE]: The skill follows security best practices by validating project configuration in CLAUDE.md and requiring user confirmation for all AI-generated content before performing write operations in Azure DevOps. No exfiltration, obfuscation, or remote code execution patterns were detected.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input (descriptions and titles) and tool outputs (parent work item data) which are interpolated into prompts for generating work item content. This risk is minimized by the skill's design, which mandates explicit user confirmation for every generated field before any modification tool is called.
Ingestion points: User-supplied arguments, description prompts, and parent work item data from the wit_get_work_item tool are used to inform generation.
Boundary markers: The workflow stops and presents generated text to the user for approval, creating a robust human-in-the-loop review boundary.
Capability inventory: The skill utilizes wit_create_work_item, wit_add_child_work_items, and wit_update_work_item to modify data within the configured Azure DevOps environment.
Sanitization: The skill applies HTML formatting to descriptions and acceptance criteria and validates that work item IDs are numeric before tool invocation.

ado-create