| name | openai-auth-switcher-public |
| description | Web-first, publishable OpenClaw skill for OpenAI OAuth account switching. Use when you need a reusable public-track workflow for first-run takeover, environment discovery, doctor checks, runtime inspection, slot management, dry-run validation, controlled switch experiments, rollback planning, and release-safe packaging without bundling live auth snapshots, logs, callbacks, or other machine-specific runtime data. |
OpenAI Auth Switcher Public
Use this skill as the publishable public-track release of the OpenAI auth switcher workflow.
It is designed for OpenClaw administrators who want a web-first, first-run-friendly, release-safe workflow for OpenAI OAuth account takeover, inspection, dry-run validation, controlled switching, and public distribution.
Purpose
Keep the live/internal operator skill and the public distributable skill separated.
This public track must:
- avoid bundling live runtime state
- avoid bundling auth snapshots, callbacks, backups, or token ledgers from a real machine
- avoid machine-specific hard-coded paths where possible
- document compatibility boundaries explicitly
- provide a release-safe packaging workflow
- keep temporary runtime files inside the skill runtime area, while encouraging important persistent state to live in an external state base via
OPENAI_AUTH_SWITCHER_PUBLIC_STATE_DIR
Core operating model
Treat OpenClaw OpenAI OAuth switching as a high-sensitivity maintenance workflow.
Always do work in this order:
- Use
install.sh as the default user-facing bootstrap entrypoint.
- Run
doctor.py when installation or environment checks fail.
- Confirm runtime discovery with
env_detect.py.
- Inspect the current runtime before any switch logic.
- Dry-run any target before proposing a write.
- Keep rollback and backup behavior explicit.
- Package only from this public skill directory or from a sanitized staging copy.
Included scripts
Primary public-release scripts:
install.sh — recommended user entrypoint; wraps the web bootstrap into a single shell commanduninstall.sh — recommended cleanup entrypoint before clawhub uninstallscripts/install_web_app.py — one-shot web bootstrap for first-run accessscripts/pick_port.py — port selection helper (9527 → 12138 → fallback)scripts/generate_web_credentials.py — default admin credential generatorscripts/doctor.py — compatibility and environment checksscripts/env_detect.py — OpenClaw path and runtime discoveryscripts/paths.py — centralized path resolution helpersscripts/inspect_runtime.py — portable runtime inspectionscripts/profile_slot.py — public-safe slot metadata and local slot filesscripts/rollback_experiment.py — rollback helper using explicit backup sourcesscripts/switch_experiment.py — controlled switch experiment with backup and rollbackscripts/token_ledger.py — local token attribution ledger rebuildscripts/hourly_usage.py — hourly/daily rollup payload for local analyticsscripts/package_public_skill.py — release-safe packager wrapper
Helper modules:
scripts/auth_file_lib.pyscripts/probe_lib.pyscripts/lock_lib.pyscripts/state_lib.py
Compatibility and safety references
Read only as needed:
references/compatibility.md — Python / Node / OpenClaw / OS expectationsreferences/runtime-discovery.md — path detection and override modelreferences/install-and-runbook.md — operator flow and first-run checksreferences/security-model.md — sensitivity, boundaries, and redaction rulesreferences/packaging-policy.md — publish checklist and forbidden contentsreferences/migration-notes.md — relation to the internal/live skill
Release rule
Do not publish skills/openai-auth-switcher directly.
Use the public skill directory for ClawHub publication and use a packaging wrapper that rejects runtime data, backups, session callbacks, and credential-bearing files.
Recommended first release positioning:
- version:
0.1.0
- tested on OpenClaw
2026.3.11
- tested on Python
3.11
- tested on Node.js
22.x
- Linux-first release
