openclaw - 💡(How to fix) Fix [Bug]: SYSTEM_RUN_DISABLED despite effective exec policy showing security=full

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

SYSTEM_RUN_DISABLED despite effective exec policy showing security=full

Steps to reproduce

Set exec security to full via config:

bashopenclaw config set tools.exec.security '"full"' openclaw config set tools.exec.ask '"off"' openclaw config set tools.exec.host '"gateway"'

Set gateway approvals to full:

bashcat > /tmp/approvals.json << 'EOF' { "version": 1, "defaults": { "security": "full", "ask": "off", "askFallback": "full" } } EOF openclaw approvals set --gateway --file /tmp/approvals.json

Restart gateway:

bashopenclaw gateway stop sleep 10 openclaw gateway start

Verify effective policy shows full:

bashopenclaw exec-policy show --json openclaw approvals get --gateway Both confirm effective security=full

Ask agent via Telegram to run any shell command:

Run this command: echo "test" Expected result: Command executes and returns "test" Actual result: Exec denied (security=deny): echo "test" SYSTEM_RUN_DISABLED: security=deny Additional context:

Bug persists across multiple gateway restarts and full stop/start cycles Reproduced on both 2026.4.26 and 2026.5.7 openclaw exec-policy show --json output:

json{"effectivePolicy":{"scopes":[{"scopeLabel":"tools.exec","security":{"requested":"full","host":"full","effective":"full","note":"requested security applies"},"ask":{"requested":"off","host":"off","effective":"off"}}]}}

tools.profile set to full gateway.nodes.denyCommands only contains camera/screen/contacts/calendar/sms — not exec

Workaround: None found. Web search used as fallback for workflows requiring shell execution.

Expected behavior

When openclaw exec-policy show --json reports effective security=full and ask=off, the agent should be able to execute shell commands without restriction. Any command sent via Telegram or a cron job session should run and return output normally.

Actual behavior

Despite the effective policy showing security=full, the agent runtime still blocks all exec-type commands with: Exec denied (security=deny): <command> SYSTEM_RUN_DISABLED: security=deny The runtime appears to be ignoring the effective policy entirely and applying a hardcoded security=deny regardless of what exec-policy show reports. The configuration layer and the enforcement layer are out of sync — config reads correctly but enforcement never updates.

OpenClaw version

2026.5.7 (eeef486)

Operating system

macOS (Apple Silicon)

Install method

No response

Model

openrouter/openrouter/free

Provider / routing chain

The exec denial occurs regardless of which model handles the request — both OpenRouter free models and Mistral return SYSTEM_RUN_DISABLED when attempting exec commands, confirming the block is at the gateway enforcement layer, not the model layer.

Additional provider/model setup details

tools.exec.security = full tools.exec.ask = off tools.exec.host = gateway exec-approvals.json defaults.security = full

openclaw exec-policy show --json confirms effective security=full But agent still receives SYSTEM_RUN_DISABLED when attempting any exec command.

Tested across: multiple gateway restarts, full stop/start cycles, versions 2026.4.26 and 2026.5.7

Logs, screenshots, and evidence

{"configPath":"/Users/otto/.openclaw/openclaw.json","approvalsPath":"/Users/otto/.openclaw/exec-approvals.json","approvalsExists":true,"effectivePolicy":{"note":"Effective exec policy is the host approvals file intersected with requested tools.exec policy.","scopes":[{"scopeLabel":"tools.exec","configPath":"tools.exec","host":{"requested":"gateway","requestedSource":"tools.exec.host"},"security":{"requested":"full","requestedSource":"tools.exec.security","host":"full","hostSource":"/Users/otto/.openclaw/exec-approvals.json defaults.security","effective":"full","note":"requested security applies"},"ask":{"requested":"off","requestedSource":"tools.exec.ask","host":"off","hostSource":"/Users/otto/.openclaw/exec-approvals.json defaults.ask","effective":"off","note":"requested ask applies"},"askFallback":{"effective":"full","source":"/Users/otto/.openclaw/exec-approvals.json defaults.askFallback"},"runtimeApprovalsSource":"local-file"}]}}

Impact and severity

No response

Additional information

No response