skills.rest
ceasarXuu
ceasarXuuCOMMUNITY
Explore all skills by this author →
GitHub

provider-switch

Install and manage reversible agent-to-model-provider profiles for Codex and Claude Code.

3|Updated Mar 20, 2026
Search Tags:#claude code#codex#installer#configuration management#deepseek#provider-switching#api credentials
One-click install
npx skills add https://github.com/ceasarXuu/AstartesSkills --skill provider-switch-ceasarxuu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: provider-switch
Source: https://github.com/ceasarXuu/AstartesSkills/tree/main/skills/provider-switch
Command: npx skills add https://github.com/ceasarXuu/AstartesSkills --skill provider-switch-ceasarxuu

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Switching a coding agent like Codex or Claude Code to a third-party model provider such as DeepSeek or StepFun normally requires hand-editing global config files, risking broken logins, leaked credentials, and irreversible changes. This Skill installs each Agent/Provider combination as an isolated, validated, and reversible side-load profile that preserves the agent's default configuration and existing login state. ## Core Features & Use Cases - Cataloged provider profiles: Ships five verified combinations (codex-deepseek-flash, codex-deepseek-pro, claude-code-deepseek, claude-code-deepseek-flash, claude-code-step) with dedicated wrapper commands like codex-ds-pro and claude-ds-flash. - Safe deterministic installers: Python scripts validate agent versions and official provider payloads, back up changed files to timestamped directories, write atomically with restrictive permissions, and preserve existing API keys without logging them. - Layered validation and rollback: Includes a loopback mock validator that checks wire-format behavior without spending provider credits, plus documented recovery steps for every profile. - Use Case: A developer wants to run Claude Code against DeepSeek V4 Pro without losing their claude.ai login. The Skill installs a claude-ds-pro command with a dedicated 0600-permission settings file, leaving global settings untouched. ## Quick Start Ask the agent to install the Codex DeepSeek Pro provider profile using provider-switch and then confirm the launch command and rollback steps.

Frequently Asked Questions about provider-switch

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I switch Claude Code to DeepSeek without losing my login?▼

Run the install_claude_deepseek.py installer, which creates a dedicated deepseek.settings.json and a claude-ds-pro wrapper command. Your global Claude settings and saved claude.ai login remain untouched, and launching claude-ds-pro routes only that session to DeepSeek.

How do I connect Codex CLI to the DeepSeek API?▼

Use the codex-deepseek-flash or codex-deepseek-pro profile installer, which downloads and validates DeepSeek's official model catalog, writes a side-load TOML config, and installs a codex-ds-flash or codex-ds-pro wrapper. Codex CLI 0.144.0 or newer is required.

Does switching providers affect my existing ChatGPT or claude.ai login?▼

No. Each profile keeps provider credentials in a separate file and never adds global login-policy fields like forced_login_method, which would invalidate an existing ChatGPT login. The installers verify login status before and after as part of validation.

Can I validate the provider setup without spending API credits?▼

Yes. The bundled validate_claude_mock.py script runs Claude Code against a local loopback mock Anthropic endpoint with a diagnostic token, checking model, effort, and auth-header behavior without any real provider request or cost.

How do I roll back a provider profile if something goes wrong?▼

Every installer backs up changed files to a timestamped provider-switch-backups directory. Restore a file by copying its backup over the managed target, or simply stop using the profile by running the ordinary agent command since global settings were never modified.

Why do the wrapper commands use bypass-permissions mode?▼

The wrappers launch with flags like --dangerously-skip-permissions as an explicit package contract, logging mode=yolo on every start. They are intended only for trusted workspaces or isolated containers, and the ordinary agent command remains unchanged with normal permission checks.

Related Skills

Explore More Software Engineering Skills→
surfingaliensurfingalien
1

codex-advisor

Queries OpenAI Codex CLI in non-interactive mode for plan reviews, code reviews, and architecture decisions.

Community
Intermediate
kinnerkarmanishkinnerkarmanish

budget

Track and enforce AI workspace costs across agents, tasks, and projects.

Community
Advanced
NiravenNiraven

api-credentials-hygiene

Audit git history for exposed API keys and verify file permissions.

Community
Intermediate
DalomeveDalomeve
3

model-route-guard

Audit openclaw.json and agent overrides to resolve model routing conflicts.

Community
Intermediate