Skip to content

Latest commit

 

History

History
72 lines (53 loc) · 1.74 KB

File metadata and controls

72 lines (53 loc) · 1.74 KB

Releasing

This repository uses lightweight, explicit releases.

Release principles

  • Prefer small, intelligible releases over large rewrites
  • Keep snapshots immutable once published
  • Update doctrine before changing public workflow claims
  • Release only after both baseline and Ralph Loop paths still validate

Versioning

Use simple semantic-ish tags:

  • v0.x.y while the project is still evolving quickly
  • bump:
    • patch for doc or small scaffold fixes
    • minor for meaningful scaffold or doctrine improvements
    • major only if the skill’s public behavior changes in a breaking way

Release checklist

  1. Validate the current skill:
python C:\Users\Seven\.codex\skills\.system\skill-creator\scripts\quick_validate.py .\skills\harness-engineer
  1. Validate any included snapshots that remain public.
  2. Compile-check the scaffold script.
  3. Smoke-test:
    • baseline scaffold
    • Ralph Loop scaffold if the preset or runner logic changed
  4. Update:
    • README.md
    • README.zh-CN.md
    • versions.json
    • ROADMAP.md if priorities changed
  5. Create a release note that explains:
    • what changed
    • whether doctrine, scaffold logic, or both changed
    • whether a new snapshot was added

When to cut a snapshot

Create a new snapshot when:

  • the scaffold behavior changes materially
  • the Ralph Loop preset meaningfully changes shape
  • a rollback point would be useful for users or maintainers

Do not create snapshots for every tiny README edit.

GitHub metadata

Keep these areas in sync with the actual repository:

  • description
  • homepage
  • topics
  • latest release notes

Recommended topics:

  • codex
  • codex-skill
  • harness-engineering
  • ai-agents
  • ralph-loop
  • scaffolding
  • long-running-agents