openclaw - 💡(How to fix) Fix [Bug]: OpenClaw v2026.03.13 OAuth with openai-codex/gpt-5.3-codex can chat but cannot execute tool actions, and modifying SOUL.md/AGENTS.md has no effect [3 comments, 3 participants]

openclaw2026-03-18 05:42:39

ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

GitHub issue URL

Helpful · Quick feedback

GitHub stats

openclaw/openclaw#49503•Fetched 2026-04-08 00:54:34

View on GitHub

Comments

Participants

Timeline

Reactions

Author

Participants

Timeline (top)

commented ×3cross-referenced ×2labeled ×2subscribed ×2

After authenticating openai-codex/gpt-5.3-codex via OAuth (with proxy to OpenAI from mainland China network), the model can chat normally but refuses to execute any tool operations, even after modifying SOUL.md/AGENTS.md as instructed.

Root Cause

RAW_BUFFERClick to expand / collapse

Bug type

Regression (worked before, now fails)

Summary

Steps to reproduce

1.Deploy OpenClaw v2026.03.13 on Ubuntu 20.04. 2.Authenticate openai-codex/gpt-5.3-codex via OAuth 2.0 flow (using a proxy to access OpenAI from mainland China network; no API key input). 3.Switch the active model to openai-codex/gpt-5.3-codex. 4.Send a tool-execution command (e.g., "Check local disk usage" to trigger df -h). 5.Observe that the bot only replies conversationally, without invoking any tool/exec action, despite claiming it has full tool permissions. 6.Modify SOUL.md and AGENTS.md to relax constraints (as suggested by the bot itself), restart the OpenClaw gateway, and repeat step 4. 7.Observe that the behavior remains unchanged: the bot still only replies with dialogue and does not execute tools.

Expected behavior

The bot should invoke corresponding tools (e.g., exec/process) to execute valid commands and return results, as it does on another identically configured server in the same LAN.

Modifications to SOUL.md/AGENTS.md should take effect and adjust the bot's execution behavior.

Actual behavior

The bot can chat normally but never triggers any tool execution for operational commands. The bot explicitly states that it is bound by multi-layer security constraints, and that SOUL.md/AGENTS.md are only one layer and cannot override higher-level system/runtime security policies. All changes to SOUL.md/AGENTS.md have no observable effect; the rules are not applied or are overridden by higher-priority system constraints. Another server in the same LAN (with identical OpenClaw version, proxy setup, and OAuth authentication method) works correctly, confirming this is not a network/proxy/OAuth configuration issue.

OpenClaw version

v2026.03.13

Operating system

Ubuntu 20.04

Install method

npm

Model

openai-codex/gpt-5.3-codex

Provider / routing chain

openclaw -> local HTTP proxy (10.10.10.223:10809) -> OpenAI API (via OAuth)

Config file / key location

~/.openclaw/openclaw.json; ~/.openclaw/auth/openai-codex.json; ~/.openclaw/SOUL.md; ~/.openclaw/AGENTS.md

Additional provider/model setup details

OAuth authentication completed successfully (no 403/401 errors), model can generate text responses normally.
Another server in the same LAN (identical version, proxy, and OAuth setup) works correctly, ruling out network/proxy/OAuth issues.
Bot explicitly states it is bound by multi-layer security constraints, where SOUL.md/AGENTS.md cannot override higher-level system policies.
Modifications to SOUL.md/AGENTS.md have no observable effect on tool execution behavior.

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response

extent analysis

Fix Plan

To resolve the issue of the bot not executing tool operations despite successful OAuth authentication and configuration modifications, follow these steps:

Update the openclaw.json configuration file to include the necessary permissions for tool execution:

{
  "models": {
    "openai-codex/gpt-5.3-codex": {
      "tool_execution": true,
      "security_constraints": {
        "override_system_policies": true
      }
    }
  }
}

Modify the SOUL.md and AGENTS.md files to relax the security constraints and allow tool execution:

# SOUL.md
allow_tool_execution: true

# AGENTS.md
tool_execution_permissions:
  - execute: true
  - override_system_policies: true

Restart the OpenClaw gateway to apply the changes:

sudo systemctl restart openclaw

Verify that the bot can execute tool operations by sending a tool-execution command (e.g., "Check local disk usage" to trigger df -h).

Verification

To verify that the fix worked, check the bot's response to tool-execution commands. If the bot successfully executes the tools and returns the results, the issue is resolved.

Extra Tips

Ensure that the openclaw.json file is properly formatted and that the tool_execution and security_constraints settings are correctly configured.
If the issue persists, check the OpenClaw logs for any error messages or security constraint violations.
Consider updating the OpenClaw version to the latest release, as newer versions may include bug fixes or security patches that address the issue.

Vote matrix · Quick signals

Works

Did the solution work? Tap to confirm.

Easy Fix

Was it a quick fix?

Time Saver

Did it save you time?

Blocking

Was it severely blocking?

Common Issue

Are others likely hitting this too?

Flaky / Intermittent

Is it intermittent?

Verified / Reproducible

Can you reproduce it reliably?

FAQ

Expected behavior

The bot should invoke corresponding tools (e.g., exec/process) to execute valid commands and return results, as it does on another identically configured server in the same LAN.

Modifications to SOUL.md/AGENTS.md should take effect and adjust the bot's execution behavior.

#api #ssr #installation #tensor shape #autograd error #output truncation #response parsing #generation error #database connection

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Data

Security

Network

Code

UI/UX

Text

System

Multimedia

Protocol

API

Engineering

openclaw - 💡(How to fix) Fix [Bug]: OpenClaw v2026.03.13 OAuth with openai-codex/gpt-5.3-codex can chat but cannot execute tool actions, and modifying SOUL.md/AGENTS.md has no effect [3 comments, 3 participants]

Recommended Tools

GitHub issue graph ai analysis

Root Cause

Bug type

Summary

Steps to reproduce

Expected behavior

Actual behavior

OpenClaw version

Operating system

Install method

Model

Provider / routing chain

Config file / key location

Additional provider/model setup details

Logs, screenshots, and evidence

Impact and severity

Additional information

extent analysis

Fix Plan

Verification

Extra Tips

FAQ

Expected behavior

Still need to ship something?

TRENDING

openclaw - 💡(How to fix) Fix [Bug]: OpenClaw v2026.03.13 OAuth with openai-codex/gpt-5.3-codex can chat but cannot execute tool actions, and modifying SOUL.md/AGENTS.md has no effect [3 comments, 3 participants]

Recommended Tools

GitHub issue graph ai analysis

Root Cause

Bug type

Summary

Steps to reproduce

Expected behavior

Actual behavior

OpenClaw version

Operating system

Install method

Model

Provider / routing chain

Config file / key location

Additional provider/model setup details

Logs, screenshots, and evidence

Impact and severity

Additional information

extent analysis

Fix Plan

Verification

Extra Tips

FAQ

Expected behavior

Still need to ship something?

RELATED_DISCOVERY

TRENDING