Init and Scaffolding

qp init helps you bootstrap quickly, either from scratch or by inferring from an existing repository.

Basic Init

Create a starter qp.yaml:

qp init

Use this for greenfield repos or when you want a minimal baseline.

Import From Existing Repo

Infer tasks from current project files:

qp init --from-repo

Current inference sources include:

• Makefile
• justfile / Justfile
• package.json
• go.mod
• Cargo.toml
• pyproject.toml, tox.ini
• pom.xml, build.gradle, build.gradle.kts
• Docker Compose files

Heuristics prioritize practical task surfaces and may mark risky flows as agent: false.

Generate Human + Agent Docs

qp init --docs

Writes:

• HUMANS.md
• AGENTS.md
• CLAUDE.md

These are derived from qp.yaml so docs stay aligned with actual task contracts.

Harness-Oriented Setup

qp init --harness

--harness focuses on operational guardrails and collaboration-ready workflows out of the box.

Codemap Bootstrap

qp init --codemap

Appends inferred codemap.packages entries (if codemap does not already exist) to accelerate explain/context quality.

Combining Flags

You can combine options for first-day setup:

qp init --from-repo --docs --codemap

Common combo by repo maturity:

1. Existing mature repo: --from-repo --docs --codemap
2. New repo with team process: --harness --docs
3. Minimal experimentation: plain qp init

Review Checklist After Init

1. Rename any awkward inferred task names.
2. Mark risky tasks with safety and agent: false as needed.
3. Add scope to key tasks.
4. Add/adjust guards.
5. Run qp validate --suggest.

Next Step

For architecture boundary enforcement in larger codebases, continue to Architecture Checks.