Unified access to Azure AI services: Search, Speech, OpenAI, and Document Intelligence. AI Search supports full-text, vector, hybrid, and semantic search with AI enrichment capabilities like entity extraction and OCR Speech service enables speech-to-text transcription (real-time and batch), text-to-speech with neural voices, speaker diarization, and custom models MCP server integration provides direct tool access via azure__search and azure__speech commands; falls back to CLI and SDK when MCP is unavailable Includes OpenAI model access, DALL-E image generation, embeddings, and Document Intelligence for form extraction and OCR
Provision OAuth 2.0 identities for AI agents with per-instance credentials and audit trails via Microsoft Graph. Creates Agent Identity Blueprints (application templates), BlueprintPrincipals (service principals), and per-instance Agent Identities, each with independent permission grants and audit scope Implements two-step fmi_path token exchange for autonomous and on-behalf-of (OBO) flows, with support for Workload Identity Federation, client secrets, and cross-tenant scenarios Provides Microsoft.Identity.Web.AgentIdentities for .NET and a containerized Microsoft Entra SDK for AgentID sidecar supporting Python, Node, Go, and Java Grants application and delegated permissions scoped per Agent Identity via appRoleAssignments and oauth2PermissionGrants ; credentials live on the Blueprint, not on individual Agent Identities
Set up tracing, logging, and monitoring for deployed ADK agents across Cloud Trace, BigQuery, and third-party platforms. Four observability tiers: Cloud Trace (always enabled, distributed tracing), Prompt-Response Logging (GenAI interactions to GCS/BigQuery), BigQuery Agent Analytics (structured agent events), and third-party integrations (AgentOps, Phoenix, MLflow, Weave, Freeplay, and others) For Agent Runtime deployments, run agents-cli infra single-project before first deploy to provision Terraform-managed infrastructure (service account, GCS bucket, BigQuery dataset); post-deployment setup requires manual IAM and env var configuration Cloud Trace works out of the box with OpenTelemetry spans tracking invocation flow, LLM calls, and tool execution; accessible via Cloud Console Trace explorer Prompt-response logging is privacy-preserving by default (metadata only via OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=NO_CONTENT ); disabled locally unless LOGS_BUCKET_NAME is set Includes troubleshooting guide covering missing traces, privacy misconfiguration, BigQuery setup, and cost optimization strategies
Generate comprehensive Product Requirements Documents that translate business vision into technical specifications. Follows a strict three-phase workflow: discovery interview to fill knowledge gaps, analysis and scoping to identify dependencies, and technical drafting using a standardized PRD schema Requires concrete, measurable success criteria and acceptance criteria; explicitly avoids vague language like "fast" or "intuitive" in favor of quantifiable benchmarks Covers executive summary, user personas and stories, technical architecture, AI system requirements with evaluation strategies, and phased rollout planning with risk analysis Includes discovery questions on core problems, success metrics, and constraints before generating any document; presents drafts for iterative feedback on specific sections
This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent…
Use Orca orchestration for structured multi-agent coordination: threaded messages, blocking ask/reply flows, task dispatch, worker_done/escalation waits, task…
Iterative evaluation and refinement patterns for improving AI agent outputs through self-critique loops. Provides three core patterns: basic reflection (self-critique loops), evaluator-optimizer (separated generation and evaluation), and code-specific test-driven refinement Supports multiple evaluation strategies including outcome-based assessment, LLM-as-judge comparison, and rubric-based scoring with weighted dimensions Includes practical Python implementations with structured JSON output parsing, iteration limits, and convergence detection to prevent infinite loops Best suited for quality-critical tasks like code generation, reports, and analysis where clear evaluation criteria and success metrics exist
Comprehensive safety analysis and improvement framework for AI prompts with detailed assessment methodologies. Evaluates prompts across eight dimensions: safety, bias detection, security, effectiveness, best practices compliance, pattern analysis, technical robustness, and performance optimization Provides structured analysis reports with risk scoring, critical issue identification, and strength assessment across all evaluation criteria Delivers improved prompt versions with specific enhancements, safety measures, bias mitigation strategies, and security hardening recommendations Includes comprehensive testing frameworks covering standard test cases, edge cases, safety testing, and bias validation with expected outcomes Offers educational insights explaining prompt engineering principles applied, common pitfalls avoided, and responsible AI best practices from industry leaders
Structured PRD generation prompt for breaking down Epics into detailed feature specifications. Generates complete Product Requirements Documents in markdown format with standardized sections covering goals, user personas, stories, functional and non-functional requirements, and acceptance criteria Prompts for clarifying questions when insufficient information is provided, ensuring comprehensive feature definition before engineering handoff Organizes output into a consistent file structure ( /docs/ways-of-work/plan/{epic-name}/{feature-name}/prd.md ) for centralized documentation management Includes explicit out-of-scope definition to prevent scope creep and establish clear feature boundaries
Interactive workflow to refine task prompts through structured questioning and clipboard delivery. Guides users through systematic prompt refinement by interrogating scope, deliverables, constraints, and technical requirements Produces polished markdown prompts and automatically copies them to the system clipboard via the Joyride extension Focuses exclusively on prompt engineering; does not generate code or implementation details Requires the Joyride extension for VSCode clipboard integration and human input requests
Polish and refine agent prompt files against proven best practices. Requires a prompt file as input; will request one if not provided Preserves front matter, encoding, and markdown structure while improving clarity and organization Corrects spelling, grammar, and wording issues without altering the original intent Applies patterns from successful prompts to strengthen structure and effectiveness
Ask an AI assistant what files it needs before answering your question. Prompts the assistant to identify required and optional files for context before attempting to answer Structures output into four categories: must-see files, helpful files, already-seen files, and remaining uncertainties Helps developers avoid incomplete answers by ensuring the assistant has examined relevant code upfront Reduces back-and-forth by clarifying dependencies and gaps in context before the actual question is answered
Deploy and manage MCP-based declarative agents across Microsoft 365 with admin center governance, role-based access, and organizational distribution. Supports five agent types: organization-published, creator-shared, Microsoft-native, external partner, and frontier agents, each with distinct deployment and approval workflows Provides admin controls for publishing, deploying to user groups, blocking, and removing agents; requires AI Admin role for full management Includes MCP-specific validation for server accessibility, OAuth authentication, tool imports, and security compliance before deployment Covers phased rollout strategies, user discovery through Copilot hub, compliance monitoring, and troubleshooting for authentication and performance issues
Query Microsoft 365 data with natural language to surface emails, meetings, documents, Teams messages, and people insights. Supports five data sources: emails, meetings, documents, Teams channels, and people/projects with natural-language prompts Install via Copilot CLI plugin (preferred) or standalone npm package; requires Microsoft 365 tenant admin consent on first use Core workflow: clarify intent, craft precise prompts with timeframe/source, run workiq ask --question "..." , and stream results Includes MCP server mode ( workiq mcp ) for exposing WorkIQ tools to other agents and workflows Best practices emphasize narrow, focused queries, privacy-respecting summaries, and mapping results to actionable follow-ups like blocking time or drafting messages
Complete development kit for building Microsoft 365 Copilot declarative agents with TypeSpec and Agents Toolkit integration. Three specialized workflows cover basic agent creation, advanced enterprise design, and validation/optimization for existing agents Supports up to 5 capabilities from 11 options including WebSearch, OneDrive/SharePoint, Graph Connectors, Power Platform, and custom connectors Enforces v1.5 schema compliance with character limits (name: 100, description: 1000, instructions: 8000) and array constraints Includes TypeSpec support for type-safe definitions, VS Code extension integration, local debugging via Agents Playground, and environment variable management