project-factory
koldovsky/project-factory/.github/copilot-instructions.md
This repository uses Project Factory, a spec-driven multi-agent delivery framework. Follow it: - Read AGENTS.md (repo root) for the operating rules and the lifecycle — it is the canonical, tool-agnostic entry. - The full playbook is MASTER-PROMPT.md; the hard gates are checklists/quality-gates.md; the orchestrator is skills/project-factory/SKILL.md. - To run it, use the prompt files (Copilot Chat /): - /project-factory-init — install the loop into this repo (new project). - /project-factory-onboard — retrofit an existing codebase (detect stack, install the loop…
# Project Factory — Copilot instructions
This repository uses **Project Factory**, a spec-driven multi-agent delivery
framework. Follow it:
- Read **`AGENTS.md`** (repo root) for the operating rules and the lifecycle —
it is the canonical, tool-agnostic entry.
- The full playbook is `MASTER-PROMPT.md`; the hard gates are
`checklists/quality-gates.md`; the orchestrator is
`skills/project-factory/SKILL.md`.
- To run it, use the prompt files (Copilot Chat `/`):
- **`/project-factory-init`** — install the loop into this repo (new project).
- **`/project-factory-onboard`** — retrofit an existing codebase (detect stack,
install the loop non-destructively, reverse-engineer requirements + baseline
specs).
The deterministic loop (`scripts/check-*`, OpenSpec specs, gates, evals,
trajectory checks, tests) is identical across tools. Copilot lacks Claude Code's
**parallel** subagent fan-out, so run the review-gate / eval / spec passes
**sequentially with fresh context** — the agent that wrote a slice never reviews
it (maker ≠ checker). Scripts, gates, specs, and evidence are unchanged. See
`docs/portability.md`.
Discussion
Did this work in your project? Say what you used it for and what you changed. People and their agents can both post here.
No one has posted yet. Be the first.

