| name | sos |
| description | Use when helping a user inspect, organize, plan, apply, activate, back up, or restore SOS-managed agent skill packs on Codex or Claude Code without assuming global installation. |
SOS Skill Orchestration
Use this skill for workflows around Skill Orchestration System (SOS) on Codex or Claude Code: inspecting active skills, proposing packs, creating reviewable plans, applying approved plans, activating packs, checking status, and restoring backups.
Default stance:
- Start in no-global-install mode.
- Use
scripts/sos_doctor.py to detect whether SOS can run and which hosts are visible.
- Resolve the host (
--host codex or --host claude) before any write command.
- Prefer read-only or dry-run commands before any write.
- Keep deterministic writes in the SOS CLI.
- Do not edit Codex config, Claude
.sos-archive/, SOS vaults, manifests, registry files, backups, or generated pointer skills by hand.
Progressive disclosure:
- Read
references/execution-modes.md when local capability is uncertain.
- Read
references/workflows.md for the specific operation the user requested.
- Read
references/safety-model.md before any write, restore, cleanup, sync apply, or source deletion.
- Read
references/codex.md for Codex-specific config or skill-root behavior.
- Read
references/claude-code.md for Claude-specific archive and skill-root behavior.
Do not load every reference up front. Do not paste the full CLI reference unless the user asks for command details or the selected workflow needs exact commands.
Keep README.md and README_CN.md aligned when user-facing install, usage, or workflow entry points change.
