name	gptprof-public
description	Public sanitized GPT profile manager for OpenClaw: switch OpenAI Codex OAuth profiles, show usage, keep the safe Pi route, and optionally try the native Codex runtime without storing secrets in the skill.
user-invocable	true
disable-model-invocation	true
command-dispatch	tool
command-tool	gptprof
command-arg-mode	raw

gptprof-public

Manage local OpenAI Codex OAuth profiles for OpenClaw, keep the safe OpenAI-Codex Pi route available, and optionally switch to the native Codex runtime route.

What It Does

/gptprof or /gptprof status shows the active profile, plan badge when available, route status, cached usage, and Telegram profile buttons.
/gptprof add starts OpenAI device authorization.
/gptprof check completes pending device authorization after the user approves it.
/gptprof refresh refreshes usage cache on demand.
/gptprof autoswitch switches only when the active profile is at or above the 95% usage threshold and a healthy spare profile is below threshold; Telegram plugin also checks autoswitch before ordinary dispatch and schedules a gateway restart after a real switch so all agent sessions reload auth.
/gptprof use-pi sets agents.defaults.model.primary to openai-codex/gpt-5.5 and agents.defaults.agentRuntime.id to pi.
/gptprof try-codex or /gptprof use-native sets agents.defaults.model.primary to openai/gpt-5.5 and agents.defaults.agentRuntime.id to codex.
/gptprof switch <slug> switches to an existing local profile unless the selected profile is already over the autoswitch threshold.
/gptt patches the current session to openai/gpt-5.5 on the native Codex runtime with the active gptprof auth profile, medium thinking, and fast mode.
/gptpro patches the current session to openai/gpt-5.5-pro with high thinking and fast mode off.

Important Routing Distinction

Auth records use provider openai-codex because the OAuth token is for ChatGPT/Codex auth. The safe fallback route is:

model: openai-codex/gpt-5.5
runtime: agentRuntime.id = "pi"

The native Codex trial route is:

model: openai/gpt-5.5
runtime: agentRuntime.id = "codex"

On OpenClaw 2026.5.3-beta.2 and newer, do not write agents.defaults.agentRuntime.fallback; config validation rejects that key for these routes.

Telegram Buttons

Buttons should be native Telegram rows:

[[{"text":"✓ profile 42%","callback_data":"gptprof:profile"}]]

Do not use abstract {label,value} buttons, do not include style, and do not pass a flat button list in this OpenClaw delivery path.

Secret Policy

This public skill contains no tokens and should never commit tokens. Runtime secrets stay in user-local files only:

~/.codex/auth.json
~/.openclaw/codex-profiles/*/auth.json
~/.openclaw/agents/*/agent/auth-*.json

The OAuth client id in the manager is public application metadata, not a client secret.

Session alias handling writes only the current OpenClaw session store entry and creates a timestamped backup before changing model override fields.

gptprof-public

Manage local OpenAI Codex OAuth profiles for OpenClaw, keep the safe OpenAI-Codex Pi route available, and optionally switch to the native Codex runtime route.

What It Does

/gptprof or /gptprof status shows the active profile, plan badge when available, route status, cached usage, and Telegram profile buttons.

/gptprof add starts OpenAI device authorization.

/gptprof check completes pending device authorization after the user approves it.

/gptprof refresh refreshes usage cache on demand.

/gptprof autoswitch switches only when the active profile is at or above the 95% usage threshold and a healthy spare profile is below threshold; Telegram plugin also checks autoswitch before ordinary dispatch and schedules a gateway restart after a real switch so all agent sessions reload auth.

/gptprof use-pi sets agents.defaults.model.primary to openai-codex/gpt-5.5 and agents.defaults.agentRuntime.id to pi.

/gptprof try-codex or /gptprof use-native sets agents.defaults.model.primary to openai/gpt-5.5 and agents.defaults.agentRuntime.id to codex.

/gptprof switch <slug> switches to an existing local profile unless the selected profile is already over the autoswitch threshold.

/gptt patches the current session to openai/gpt-5.5 on the native Codex runtime with the active gptprof auth profile, medium thinking, and fast mode.

/gptpro patches the current session to openai/gpt-5.5-pro with high thinking and fast mode off.

Important Routing Distinction

Auth records use provider openai-codex because the OAuth token is for ChatGPT/Codex auth. The safe fallback route is:

model: openai-codex/gpt-5.5

runtime: agentRuntime.id = "pi"

The native Codex trial route is:

model: openai/gpt-5.5

runtime: agentRuntime.id = "codex"

On OpenClaw 2026.5.3-beta.2 and newer, do not write agents.defaults.agentRuntime.fallback; config validation rejects that key for these routes.

Telegram Buttons

Buttons should be native Telegram rows:

[[{"text":"✓ profile 42%","callback_data":"gptprof:profile"}]]

Do not use abstract {label,value} buttons, do not include style, and do not pass a flat button list in this OpenClaw delivery path.

Secret Policy

This public skill contains no tokens and should never commit tokens. Runtime secrets stay in user-local files only:

~/.codex/auth.json

~/.openclaw/codex-profiles/*/auth.json

~/.openclaw/agents/*/agent/auth-*.json

The OAuth client id in the manager is public application metadata, not a client secret.

Session alias handling writes only the current OpenClaw session store entry and creates a timestamped backup before changing model override fields.