From azure-devops-agents-claude
Use when the user wants to plan, setup, build, design, define, implement, secure, expose, document, diagram, estimate, schedule, or break down a software, platform, infrastructure, DevOps, security, UX, or technical documentation initiative that should become RUP-style Azure DevOps artifacts.
How this skill is triggered — by the user, by Claude, or both
Slash command
/azure-devops-agents-claude:azure-devops-planningThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill when the user expresses planning intent for software, platform, infrastructure, DevOps, security, UX, or documentation work. The user does not need to mention Azure DevOps, RUP, or a route name.
Use this skill when the user expresses planning intent for software, platform, infrastructure, DevOps, security, UX, or documentation work. The user does not need to mention Azure DevOps, RUP, or a route name.
Treat natural language intent as a planning trigger when it starts or implies work such as "I want to", "we need to", "we should", "let's", "setup", "build", "design", "create", "define", "implement", "secure", "expose", "integrate", "document", "diagram", "estimate", or "break down".
Examples that should trigger:
If the request is not already phrased in RUP terms, map it to the closest canonical concept before planning. Ask a clarifying question only when the target concept or scope is genuinely ambiguous.
Plan in RUP-style SDLC language:
Map to Azure DevOps process terms only after retrieving and validating the target process metadata with mcp_ado_wit_list_backlogs and mcp_ado_wit_get_work_item_type.
Preferred:
/capture-request/define-requirements/design-ux/plan-requirement/document-solution/plan-delivery/plan-taskRun the SDLC role sequence in order, but stop at the last phase the request needs. For each phase, produce only that phase's output and pause before continuing.
Pause after each phase. Never create Azure DevOps work items until the user confirms the final plan.
If the user asks to build, implement, change, fix, configure, deploy, secure, expose, integrate, document, diagram, or otherwise produce project work, first verify that the request is traceable to an existing approved Azure DevOps work item or confirmed RUP planning artifact.
If the work is not already represented in Azure DevOps, do not start development. Trigger planning by capturing the request as a new Stakeholder Request or Change Request, then run the SDLC phases until ready. Development can start only after requirements, UX for user-facing work, architecture, technical documentation, delivery planning, and needed implementation tasks are confirmed.
Architecture is not a technology list. It must define the system boundary, actors, components, runtime flows, deployment, data, security, operations, decisions, tradeoffs, and open questions. Every technology choice must be confirmed or listed as an ADR candidate.
Technical documentation must not introduce architecture decisions. Mermaid diagrams for Azure DevOps wiki must use ::: mermaid blocks, graph TD; or graph LR; for flowcharts, simple node IDs, quoted ASCII labels, no HTML, no Markdown labels, no angle-bracket placeholders, no raw Unicode symbols, and no GitHub-style Mermaid code fences.
Use the official azure-devops MCP server and mcp_ado_* tool names.
Before writing, call mcp_ado_wit_list_backlogs and mcp_ado_wit_get_work_item_type to build the process profile. Create with mcp_ado_wit_add_child_work_items, update only discovered fields with mcp_ado_wit_update_work_item, and verify links with mcp_ado_wit_get_work_item.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub sarins-lab/azure-devops-agents --plugin azure-devops-agents-claude