sol-luna-setup
yuri-nagasaki/subagent-skills · skills.sh
Open source Repository Open in the app JSON README (API)
About
Skill publicada por yuri-nagasaki/subagent-skills no skills.sh. Instale com: npx skills add yuri-nagasaki/subagent-skills@sol-luna-setup
Details
- Kind
- Agent skills
- Publisher
- yuri-nagasaki
- Origin
- skillssh
- Category
- ferramentas
- Stars
- 12
- Forks
- 1
- Last push
- 2026-08-02T11:20:53Z
- Repository state
- ativo
- Language
- Shell
- License
- MIT
- Added
- 2026-10-07 05:37:48
- Updated
- 2026-10-07 05:37:48
- Origin id
yuri-nagasaki/subagent-skills/sol-luna-setup
README
# subagent-skills Project-level templates and a setup skill for **Sol (leader) + Luna (workers)** multi-agent workflows in Codex, Claude Code, and Pi. Blog (Chinese, longer write-up): https://catcat.blog/2026/08/sol-luna-layered-subagents-codex-claude-pi.html ## Install ```bash npx skills add Yuri-NagaSaki/subagent-skills -g -y ``` Claude Code only: ```bash npx skills add Yuri-NagaSaki/subagent-skills -g -a claude-code -y ``` Project-local: ```bash npx skills add Yuri-NagaSaki/subagent-skills -y ``` List skills in this repo: ```bash npx skills add Yuri-NagaSaki/subagent-skills -l ``` Installed skill: **`sol-luna-setup`**. After install, ask the agent to run that skill on the current project. ## Optional: write project files ```bash git clone https://github.com/Yuri-NagaSaki/subagent-skills.git export OPENAI_API_KEY=... # env only; never commit bash subagent-skills/scripts/bootstrap.sh /path/to/project ``` Creates or merges: - `.codex/config.toml`, `.codex/agents/luna_*.toml` - `AGENTS.md` - `.claude/agents/luna-*.md`, `CLAUDE.md` - `scripts/prepare-luna-catalog.sh` Edit `base_url` in `.codex/config.toml` for your provider. ## Layout ```text skills/sol-luna-setup/ # discovered by npx skills templates/ # codex / claude / pi copies scripts/bootstrap.sh scripts/prepare-luna-catalog.sh demo/ # tiny auth sample + npm test docs/sol-luna-catalog-fix.md ``` ## Sol cannot spawn Luna Common error: ```text Unknown model `gpt-5.6-luna` for spawn_agent ``` ```bash bash scripts/prepare-luna-catalog.sh "$(pwd)/.codex/models-v1.json" ``` Point `model_catalog_json` at that file. Keep `multi_agent_v2 = false`. Details: [docs/sol-luna-catalog-fix.md](https://github.com/yuri-nagasaki/subagent-skills/blob/HEAD/docs/sol-luna-catalog-fix.md). ## Smoke checks ```bash codex exec --sandbox read-only -c 'model="gpt-5.6-sol"' \ "Reply: SOL_SMOKE_OK" </dev/null codex exec --sandbox read-only -c 'model="gpt-5.6-luna"' \ "Reply: LUNA_SMOKE_OK" </dev/null codex exec --sandbox read-only \ "Use luna_scout per AGENTS.md; start answer with SCOUT_DONE" </dev/null ``` `codex exec` reads stdin; always redirect `</dev/null` in scripts. Demo: ```bash cd demo && npm test ``` ## Roles | Role | Model | Notes | |------|--------|--------| | Main session | `gpt-5.6-sol` | Plan, review, commit/PR | | luna_scout | `gpt-5.6-luna` | Read-only | | luna_worker | `gpt-5.6-luna` | Bounded writes; no commit | | luna_critic | `gpt-5.6-luna` | Adversarial review | | luna_tester | `gpt-5.6-luna` | Run specified tests | ## Safety - Do not commit API keys, host credentials, or private IPs. - Use env vars (`OPENAI_API_KEY`, `GATEWAY_API_KEY`). - Ignore `models-v1.json` in git (generated locally). - Demo auth code is not production-safe. ## License [MIT](https://github.com/yuri-nagasaki/subagent-skills/blob/HEAD/LICENSE)