ATP Agent Skills

On this page

Before you start
One-command install
Included skills
Verify the install
Next steps

ATP Agent Skills are public operating guides for Codex, Claude Code, and other agents that support SKILL.md. They teach an agent ATP authentication, unified model names, request payloads, asynchronous media tasks, and common errors without exposing provider-specific APIs in your application.

Before you start#

Skills are the layer above the connection, not a replacement for it. Point the agent at the Gateway first, then install these.

Connect your agent first

Base URL, key, and model for Claude Code, Codex CLI, and every other agent — plus the three tools that need special handling.

Skills never store your API key

Skills are Markdown instructions only. Keep the atp- key in an environment variable or secret manager; never put it in SKILL.md, source code, or git.

One-command install#

Install all seven official skills:

Shell

# Codex
curl -fsSL https://atptoken.ai/skills/install.sh | sh -s -- codex

# Claude Code
curl -fsSL https://atptoken.ai/skills/install.sh | sh -s -- claude

# Install for both
curl -fsSL https://atptoken.ai/skills/install.sh | sh -s -- both

The installer does not overwrite existing files by default. Explicitly allow replacement when updating:

Shell

curl -fsSL https://atptoken.ai/skills/install.sh | ATP_SKILLS_FORCE=1 sh -s -- codex

If your security policy disallows piping into a shell, download, inspect, and run:

Shell

curl -fsSL https://atptoken.ai/skills/install.sh -o /tmp/install-atptoken-skills.sh
less /tmp/install-atptoken-skills.sh
sh /tmp/install-atptoken-skills.sh codex

Included skills#

Skill	Use it for
`atptoken-gateway`	Authentication, model discovery, shared errors, files, and media routing
`atptoken-openai`	OpenAI-compatible chat and Responses-style workflows
`atptoken-anthropic`	Anthropic Messages format and Claude tools
`atptoken-gemini`	Native Gemini `generateContent` format
`atptoken-image`	Synchronous image generation
`atptoken-video`	Text/image-to-video, task creation, and polling
`atptoken-audio`	TTS, speech, and audio tasks

Direct downloads: Gateway · OpenAI · Anthropic · Gemini · Image · Video · Audio.

The machine-readable catalogue is available at /skills/manifest.json.

Verify the install#

Shell

# Codex
find "${CODEX_HOME:-$HOME/.codex}/skills" -maxdepth 2 -name SKILL.md \
  -path "*/atptoken-*"

# Claude Code
find "${CLAUDE_HOME:-$HOME/.claude}/skills" -maxdepth 2 -name SKILL.md \
  -path "*/atptoken-*"

Then ask the agent to “generate an image with ATP” or “create a Wan video task with ATP.” It should first call GET https://api.atptoken.ai/v1/models with the same project key, then choose a model enabled for that project — the same opening move described in Connect any coding agent. If the agent skips it and hard-codes a model name instead, that is the behaviour to correct.

Project permissions still apply

Installing a skill does not enable a model or bypass allowed models. If a model is absent from GET /v1/models, it cannot be used; enable it for the project in Console Resources first.

Next steps

PreviousCoding agents NextOpenAI SDK

Was this page helpful?