Start here
Getting Started
Prerequisites, installing the package with the correct target, and your first run.
InstallRun the Squad
Profiles, autonomy modes, remote approval, and how the coordinator routes work.
ConsumerDemo
Two runnable walkthroughs — an Azure squad (cost, architecture, IaC, gated deploy, as-built, diagnose) and a non-technical product squad (discovery, requirements, roadmap, deck) — each with a say-it-once autopilot version.
WalkthroughEcosystem
The MCP server and the Copilot CLI plugin — what each surface is for, and where their repositories and documentation sites live.
ReferenceMaintaining
Dependency generation, the author workflow, customization, and the release process.
AuthorTroubleshooting
Known install errors with copy-paste fixes, versioning, and repository notes.
ReferenceWhat this project does
This repository provides a curated APM package that references content from
microsoft/hve-core under
.github/agents, .github/instructions,
.github/prompts, and .github/skills. Instead of hand-maintaining
hundreds of dependency paths, an automation script regenerates the dependency list in
apm.yml.
One bundle
Auto-discovers supported markdown artifacts from selected HVE Core folders and ships them together.
Reproducible
Dependencies are resolved into apm.lock.yaml so installs are repeatable.
Squad included
A locally authored Squad Coordinator plus supporting agents, instructions, a prompt, and a skill.
Azure execution layer
Architecture, IaC, deployment, post-deploy as-built documentation, and resource triage — all from the squad. Deploys stay behind a human gate and the new Azure MCP capability enables live governance discovery.
Modernization
A squad role that plans framework and dependency upgrades and cross-stack re-platforms, then routes execution to the developer role or the official App Modernization tooling.
Federation
Run several named sub-squads in one repository — a product squad for the business team, an azure squad for architects — each an ordinary squad under its own state root, routed by /squad-federation. Opt-in and additive.
Discovery gate
When a run has nothing written down yet, the squad offers a brainstorm before it builds — framing, options, and a short brief the intake gate then validates. Offered rather than automatic, because ideation is your ideas, not the model's. Ships with the product and full profiles.
Requirements intake gate
Before planning or building on a PRD, BRD, spec, or other input, the squad checks it for completeness and clarity, records a readiness verdict, and auto-remediates blocking gaps before proceeding. Ships with the product and full profiles.
Companion projects
Beyond the local /squad experience, the squad reaches two other hosts through
sibling repositories. Neither is a fork — both are built from a tagged
hve-squad release, so each version maps back to a known version of this package.
See Ecosystem for how to choose between them.
MCP server
hve-squad-mcp publishes the squad as model-invocable tools —
squad_research, squad_plan, squad_review,
squad_architect, squad_run, and a deterministic
squad_render_pptx file-output tool — over Streamable HTTP with Entra
auth, so hosts such as Copilot Studio can call the squad.
Copilot CLI plugin
hve-squad-plugin is a generated, release-gated mirror of the squad's plugin
tree, published as a Copilot plugin marketplace. It runs the squad in the GitHub Copilot
CLI and the desktop app with no apm install.
Ecosystem overview → hve-squad-mcp docs → hve-squad-plugin docs →
Quick install
# In the project you want the squad in:
apm install "Peter-N91/hve-squad#vX.Y.z" --target copilot
Then invoke the coordinator from Copilot Chat with the /squad prompt. See
Getting Started for the full flow.