copilot-studio-build
Skill1 starsChanged 2 months agoMITView on GitHub
Build a Microsoft Copilot Studio agent — declarative agent JSON, knowledge sources, actions/plugins, publish to Teams/M365.
In this repository33 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- decideSkill
- deliverSkill
- researchSkill
- sprintSkill
- superteam-coachSkill
- superteam-coachSkill
- azure-ai-foundry-devSkill
- azure-devopsSkill
- bc-mcpSkill
- decideSkill
- deliverSkill
- erp-connectorsSkill
- fabric-appsSkill
- fabric-medallionSkill
- fabric-notebooksSkill
- fabric-onelakeSkill
- fabric-power-biSkill
- find-skillsSkill
- mirrorSkill
- ontologySkill
- playwrightSkill
- researchSkill
- sap-btp-agentsSkill
- semantic-reasoningSkill
- semantic-searchSkill
- sprintSkill
- superteam-coachSkill
- superteam-demoSkill
- systematic-debuggingSkill
- test-driven-developmentSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
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
- Distinctive: line 23
Test: Run the acceptance suite: 10+ prompts covering happy paths, edge cases, refusal cases (topics the agent must refuse).
- Distinctive: line 32
Verify the declarative agent schema version at author time — never assume the last one still applies
- Distinctive: line 33
Every knowledge source has explicit scope — no tenant-wide "search everything" grounding
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.
Files
/copilot-studio-build — Custom Copilot Authoring
Copilot Studio ships custom agents into Teams, M365 Copilot, and standalone chat channels. A production agent is a JSON manifest (declarative agent format) plus knowledge sources, actions, and a publish channel. This skill goes from brief to published in one loop.
When to invoke
- Steve or Superman routes work to build a Copilot Studio agent for a client
- Nexus is standing up a Microsoft platform demo that needs a bespoke copilot
- An existing agent needs a new knowledge source or action wired in
- A publish channel (Teams, M365 Copilot, Web) needs configuring
Steps
- Agent brief. Capture: agent name, one-line purpose, target user, channels, tone-of-voice, knowledge sources, actions available, guardrails (topics to refuse). Sign off with the stakeholder before authoring.
- Declarative agent JSON. Author the manifest against the current declarative agent schema — verify version at author time (schema versions have shifted through 2025-2026). Fields include
name,description,instructions,conversation_starters,capabilities,actions. Athanasios verifies the schema URI before publish. - Connect knowledge sources. SharePoint sites, OneDrive files, Dataverse tables, public web (allowlisted domains), or Graph connectors. Scope narrowly — over-broad grounding produces noisy answers. Test grounding with 5+ representative questions.
- Wire actions and plugins. Actions are the agent's tools — REST connectors, Power Automate flows, or Copilot Studio topic-based logic. Each action has an OpenAPI spec or connector definition. Enforce least privilege on connections.
- Publish to Teams / M365 / Web. Configure channel per target. Teams requires app manifest and admin approval. M365 Copilot embedding requires tenant admin allow-listing. Web deployment requires a Direct Line channel.
- Test. Run the acceptance suite: 10+ prompts covering happy paths, edge cases, refusal cases (topics the agent must refuse). Log outcomes. Iterate until every prompt passes.
Deliverable format
agent.json— declarative agent manifest committed to Gitknowledge-sources.md— list of sources with scope and refresh cadenceactions/*.yaml— one OpenAPI or connector spec per actiontest-plan.md— the acceptance-suite prompts and pass/fail results
Non-negotiables
- Verify the declarative agent schema version at author time — never assume the last one still applies
- Every knowledge source has explicit scope — no tenant-wide "search everything" grounding
- Refusal-topic tests are part of every acceptance suite
- Publish only after tenant admin has approved (Teams / M365 Copilot channels)
References
- Copilot Studio — overview
- Declarative agents for Microsoft 365 Copilot
- Copilot Studio — actions and connectors
- Publish to Teams from Copilot Studio
- Superteam ADR — MS Fabric skills pack
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
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.