>_ The Manifest
~/collections $ cat best-practices.mdx

Best Practices

Curated learning path • 7 posts

Write better instructions, design engaging conversation starters, and steer model behavior so users love interacting with your declarative agent.

Building a great declarative agent is about more than just connecting data sources. This collection covers the craft of writing effective instructions, going beyond the basics with advanced patterns for persona, scope, and error handling, reserving exact editorial answers for known questions, designing conversation starters that guide users to value (including conditional starters that appear only when their capabilities are configured), and steering model behavior with behavior overrides and the default response mode to reduce hallucination and set the right tone from the first message.

  1. 1
    · 7 min

    Crafting Effective Instructions for Declarative Agents

    Master the 5-part instruction framework that turns a generic declarative agent into a purpose-built AI assistant, with real examples from Zava Insurance.

    #declarative-agents #m365 #copilot
  2. 2
    · 12 min

    Advanced Instruction Engineering for Declarative Agents

    Go beyond basics with advanced patterns for persona engineering, scope boundaries, error handling, and iterative testing of declarative agent instructions.

    #copilot #declarative-agents #m365
  3. 3
    · 7 min

    Editorial Answers Make Declarative Agents Predictable

    Editorial answers in declarative agent manifest v1.7 let you reserve exact responses for known questions, with semantic matching catching natural language variations.

    #copilot #declarative-agents #extensibility
  4. 4
    · 6 min

    Designing Conversation Starters for Declarative Agents

    Create conversation starters that showcase your declarative agent's capabilities and guide users to high-value interactions from the very first message.

    #declarative-agents #m365 #copilot
  5. 5
    · 8 min

    Conditional Conversation Starters for Declarative Agents

    Manifest schema v1.7 adds depends_on to conversation starters, so prompts only show up when the capabilities they need are configured.

    #copilot #declarative-agents #extensibility
  6. 6
    · 5 min

    Behavior Overrides in M365 Copilot Declarative Agents

    Configure behavior overrides in declarative agents to prioritize knowledge sources, reduce hallucination, and control follow-up suggestions in M365 Copilot.

    #declarative-agents #m365 #copilot
  7. 7
    · 6 min

    Set a Default Response Mode for Your Declarative Agent

    Use the new default_response_mode manifest property to control whether your declarative agent starts in Auto, Quick response, or Think deeper mode.

    #declarative-agents #copilot #behavior-overrides