QRSPI for Codex

Codex-native multi-agent orchestration for software development using the QRSPI workflow.

This repository is designed to be the central source of truth for a reusable Codex skill plus a thin per-project stub. It is intentionally language-agnostic: QRSPI should choose the most suitable language and tooling for the task and existing repository, not assume Python.

Repository contents

• skills/qrspi-codex-orchestrator/ — installable Codex skill
• examples/project-stub/ — minimal AGENTS.md stub for any target repository
• scripts/build_release.py — builds release-ready ZIP artifacts from this repository
• docs/INSTALL.md — installation and usage guidance
• docs/PUBLISH.md — GitHub publishing and release guidance
• .github/workflows/ — CI and release automation

Why this repository exists

QRSPI should be reusable across many repositories without copying a full starter repo every time.

The intended model is:

1. Install the skill once in Codex.
2. Add only a thin AGENTS.md stub to each target repository.
3. Let the skill bootstrap .qrspi/ on demand when a task starts.

Install in Codex

Use the Codex skill installer with the GitHub directory URL of the skill folder:

$skill-installer install https://github.com/<owner>/<repo>/tree/main/skills/qrspi-codex-orchestrator

Then restart Codex.

Detailed installation instructions are in docs/INSTALL.md.

Use in an existing repository

1. Copy or adapt examples/project-stub/AGENTS.md into the target repository root.
2. Start Codex in the target repository.
3. Ask Codex to use QRSPI.
4. If .qrspi/ is missing, let the skill bootstrap it on demand.

Example prompts:

QRSPI start: implement feature X and stop after design for approval.

QRSPI resume.

QRSPI status.

Design principles

• explicit phase order: question -> research -> design -> structure -> plan -> implement -> pr
• human gate after design
• one vertical slice at a time
• minimal per-project footprint
• choose language and tooling from evidence:
  1. explicit user requirement
  2. repository conventions
  3. lowest-friction implementation

Publish this repository on GitHub

This repository is already structured for direct publication:

• root LICENSE
• repository README
• install and publish docs
• release build script
• CI workflow
• tag-based release workflow

See docs/PUBLISH.md for the exact sequence.

Build release artifacts locally

python scripts/build_release.py

This creates:

• dist/skill.zip
• dist/qrspi-codex-native-repo.zip

License

• Repository: MIT, see LICENSE
• Skill-specific license file: skills/qrspi-codex-orchestrator/LICENSE.txt