Boost OS for GitHub Copilot

Structure for Copilot's coding agent

Copilot's agent follows repo instructions, and Boost OS gives it a real instruction set: an AGENTS.md kernel with rules like “debug from evidence” and “preserve user changes,” plus 20 commands it activates when you type the slash.

Because everything is a plain file in the repo, it works across Copilot's surfaces and for every teammate who clones the project. Copilot doesn't expose a hook system, so the always-on guards stay dormant here. The commands, kernel, and memory work in full.

How it installs in GitHub Copilot

Paste one prompt into GitHub Copilot. No terminal.

Click the onboarding pill, paste one prompt into GitHub Copilot
The kernel installs to AGENTS.md, which Copilot's coding agent reads for repo instructions
Commands install to .agents/skills/. Type /os in the Copilot chat to see the menu
Project memory scaffolds in .agents/memory/ and ships with the repo
Guards need host hook support, which Copilot doesn't expose. Every command still works

Where GitHub Copilot users start

The full menu

Every command below installs with that one prompt and appears in your GitHub Copilot chat the moment you type the slash. Browse all 20 commands.

/os: list every Boost OS slash command available in this project, grouped by category
/cost: break down a feature or vendor across money, build time, learning curve, lock-in, and ongoing maintenance
/design: shape a vague feature idea into a plan you can hand to /preflight before any code gets written
/explain: translate the previous output into plain language for a non-coder
/next: draft 2-3 builder-friendly prompts you could send next, based on the recent chat
/spawn: fan a task into parallel subagents when the work has independent parts
/audit: find legacy and dead code in a bounded scope, then delete with per-file confirmation
/health: check production error logs for a service in your stack and surface a grouped report
/mobile: audit one screen for mobile readiness and hand back a numbered plan you can give to /preflight
/preflight: audit a plan against the existing codebase to catch drift, bloat, redundancy, and misalignment before you start coding
/security: pre-deploy safety sweep across secrets, env hygiene, dependency advisories, and risky API routes
/sync: sync the agent's knowledge of installed library versions so it stops writing against outdated APIs
/vibe: sanity-check your uncommitted work and tell you whether you're on track or about to ship a problem
/forget: delete a saved subsystem reference
/recall: read a saved subsystem reference and pull its contents into the current chat
/remember: capture what this session learned about a subsystem into a reference file
/new: build your own custom slash command for this project through a short conversation
/pause: turn off the always-on parts of Boost OS without removing anything; /resume brings them back
/resume: turn the always-on parts of Boost OS back on after a pause
/uninstall: cleanly remove Boost OS from the project, including managed skills, project memory, and references (your own commands stay)
/update: check for and install Boost OS updates from any AI chat: newer commands, kernel improvements, fixes
/whats-new: see what shipped since your installed version, without installing anything

Also works in

Install Boost OS in GitHub Copilot today

Install it once. Your agent gets production code to build from, twenty-two commands that catch mistakes, and a memory of your project. One license, lifetime updates.

From $99 · 7-day refund · license key emailed instantly