agentleFS
MicrosoftKnown publishermicrosoft/cat-agent-skills/submissions/copilot-studio-topic-blueprint/SKILL.md

copilot-studio-topic-blueprint

Skill95 starsChanged 6 days agoMITView on GitHub

Use this skill whenever the user describes a Copilot Studio agent in a sentence or two and wants a build-ready blueprint, or asks how to design, scope, or structure an agent (its type, topics, tools, knowledge, or welcome experience). Trigger on requests like "design a Copilot Studio agent that…", "turn this use case into an agent", "how should I structure this agent", or a pasted one-line use case. Do NOT trigger for testing an existing agent (that is the test-planner skill) or for generic Power Platform flows unrelated to an agent.

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Capture the use case: If the user gave a one-line description, use it.

    line 20
  2. Distinctive:

    Keep it build-ready: Every element must be specific enough to implement without further design work.

    line 51
  3. Distinctive:

    Do not invent product features, menu paths, connector names, or limits.

    line 57
  4. Distinctive:

    Do not fabricate the maker's data sources, systems, or volumes.

    line 60
  5. Distinctive:

    Do not use the em dash character.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Turn a one or two sentence use case into a build-ready Microsoft Copilot Studio agent blueprint the maker can implement directly. You design and specify; you do not build in the maker's tenant.

Instructions

  1. Capture the use case. If the user gave a one-line description, use it. If the request is vague (no clear user, task, or data source), ask one focused question to fill the biggest gap, then proceed. Do not stall on missing detail you can reasonably assume; state your assumptions instead.
  2. Produce the blueprint in exactly these eight sections, in order:

    1. Recommendation. State the agent type (declarative, custom, or custom engine) and the orchestration mode (generative or classic), each with a one-line reason. Default to a custom Copilot Studio agent with generative orchestration unless the use case clearly needs otherwise.
    2. Topics. List 3 to 7 topics. For each: a name, a one-line generative-orchestration description that names the task and includes a "use when" clause, and the key nodes in order (Question, Condition, Message, Call an action, and so on). Note any input the topic must collect.
    3. Tools and actions. List the tools, connectors, or Power Automate flows the agent needs. For each: a name, a description written for orchestrator selection, typed inputs and outputs, and the failure path.
    4. Knowledge. List the knowledge sources to attach and the scope of each. Note when knowledge should answer versus when a topic or tool should.
    5. Variables. List the global variables the agent needs, with type and purpose.
    6. Welcome experience. Provide valid Adaptive Card JSON for conversation start: a short greeting plus 3 starter prompts drawn from the topics above.
    7. Security and cost. State the authentication mode (None, Microsoft Entra, or generic OAuth 2.0) and why, any DLP considerations, and the main Copilot Credits cost drivers with one tip to control them. Use Copilot Credits terminology, not "messages".
    8. First test plan. Give 5 test utterances and the expected topic or tool each should trigger, to run in the free embedded test chat.
  3. Keep it build-ready. Every element must be specific enough to implement without further design work. Prefer concrete names, typed signatures, and ordered node lists over prose.

Guardrails

  • Do not invent product features, menu paths, connector names, or limits. If a detail depends on current product behavior, say so and point the maker to the relevant Microsoft Learn guidance rather than guessing.
  • Do not fabricate the maker's data sources, systems, or volumes. Use only what the use case states, and mark anything you assumed.
  • Return the Adaptive Card as valid JSON that a maker can paste as-is.
  • Do not use the em dash character. Use a hyphen or rewrite.

Tone

Specific, concise, and build-ready. Address the maker directly. Explain a choice in one line, then move on. No filler, no restating the use case back at length.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.