English · 简体中文
Run independent Codex Desktop work in the same local project with a small pool of verified Luna tasks. Sol assigns exclusive write paths, checks each result, and reuses a completed task for the next related unit.
How it works · When to use it · Quick install · Compatibility · Validation · Contact
Suppose Luna A handles supplier content while Luna B handles a separate UI unit. When A finishes, Sol checks its changes and gives A the next related content unit with a new write allowlist. B can continue its own work. The next unit does not require a third conversation.
- Reuse verified workers first: match the exact saved project and confirm the task's origin, model, reasoning effort, local environment, and completed state. A matching title alone is not proof.
- Keep ownership explicit: each assignment gets its own exclusive write paths and validation. The previous allowlist expires after Sol accepts the result.
- Accept the integrated result in Sol: wait for every assigned task, review changed paths, then run combined verification. Luna's completion report is a handoff.
The Skill instructions define the full dispatch gate, boundaries, and prompt contract.
Use it when a Sol-led Codex Desktop task has at least two independent units with non-overlapping write paths in the same saved local project. The skill may also be invoked explicitly when you ask to delegate parallel work to Luna; its dispatch gate still applies. Keep coupled or shared-file work with Sol.
Give this instruction to an Agent with command-line access:
Install sol-luna-local-parallel: https://raw.githubusercontent.com/Fable-Forge/sol-luna-local-parallel/main/docs/install.md
Or use the Agent Skills CLI:
npx skills add Fable-Forge/sol-luna-local-parallelRead the installation guide first. See update and uninstall for lifecycle instructions.
- Supported: Codex Desktop with local project tasks and
gpt-6-luna/xhigh - Maturity:
experimental - GitHub Topics:
codex,multi-agent,parallel,workflow
Installation does not guarantee that an already-running session will hot-load the skill. Verify a natural-language trigger in a fresh session after installation. An exact saved-project match and two safe concurrent work units are required before dispatch.
SKILL.md: triggers, boundaries, and primary workflowagents/openai.yaml: Codex display metadatareferences/: detailed material loaded on demand, when presentscripts/: reusable tools and repository validator, when presentdocs/: install, update, and uninstall guides
Structural validation, installation visibility, real triggering, and final output quality are separate claims. Passing CI proves only the repository structure and static rules; a real Agent trigger still requires its own acceptance test.
If this skill helps your workflow, star the repository so other Codex Desktop users can find it.
- 📚 All skills: FableForge Agent Skills
- 📧 Email: 53815263@qq.com
- 🐛 Bugs and feature requests: sol-luna-local-parallel Issues
- 💬 Usage questions and business collaboration: FableForge Discussions or email
MIT © 2026 FableForge