OpenClaw tools.exec.security_full_configured: Posture Warning or Real Problem?
Published 2026-10-20 · Updated 2026-10-20 · By The Is My Agent Safe team
Short answer
Your agent can run host commands without asking first (exec security is full). OpenClaw calls that a posture warning, not proof of a bug. It turns critical when the audit also finds open channel access. Fix it first if others can message your agent: close the open channel, then switch exec to the cautious preset so commands not on your allowlist ask a human.
- OpenClaw rates it warn, or critical when the audit also detects open DM or group access.
- OpenClaw's docs say security="full" alone "is a broad posture warning, not proof of a bug" and is the chosen default for trusted-operator setups.
- To require approvals: stop the Gateway, run openclaw exec-policy preset cautious, start the Gateway and re-run the audit.
- If the finding is critical, close the open channel first. That clears the critical part.
Paste your audit: run openclaw security audit --deep and paste the output into our free check for a red/amber/green result with the most important finding first (Limits of the check).
What should I do first?
Look at the severity on the finding.
- Critical: the detail also lists open channel access, such as
channels.telegram.dmPolicy. Close those first. Set each listed DM policy topairingorallowlist, and each group policy toallowlist. - Warn: only exec security is flagged. OpenClaw treats this as a posture choice. If you're the only person who can message the agent, you can keep it. If not, tighten it.
To make commands that aren't on your allowlist ask a human first:
openclaw exec-policy show openclaw gateway stop openclaw exec-policy preset cautious openclaw gateway start openclaw security audit --deep
OpenClaw's exec approvals docs describe cautious as host=gateway, security=allowlist, ask=on-miss, askFallback=deny. The same docs say preset and set refuse to change policy while a live Gateway owns the state, so "stop a running Gateway through its service owner" first. That's why the commands stop it.
Presets set host as well, and cautious sets host=gateway. If your agents run exec in a sandbox, set only the fields you want instead, for example openclaw exec-policy set --security allowlist --ask on-miss. The CLI docs say set "applies only the flags you pass". For exec on a node, the same docs say to use openclaw approvals set --node <id|name|ip>, because exec-policy can't change node approvals.
What changes after you tighten it?
OpenClaw's exec docs describe the modes:
| Mode | What happens |
|---|---|
full | "No ordinary policy prompts" |
ask | "Allowlist matches run directly; everything else asks a human." |
allowlist | Only allowlisted commands run, and nothing else is asked |
deny | "Exec is denied." |
Commands the agent used to run straight away may now wait for your approval. Allow the ones you trust, as described in the exec approvals docs. To go back to no-approval host exec, the same docs show openclaw exec-policy preset yolo.
What does tools.exec.security_full_configured mean?
OpenClaw's security audit checks page lists it as warn/critical: "Host exec is running with security="full"". The settings it points to are tools.exec.security and agents.entries.*.tools.exec.security.
In OpenClaw 2026.10.5's source, the finding appears when exec security works out to full for the default agent or any named agent, from either tools.exec.security or tools.exec.mode. It's critical when the audit also finds open DM or group access. Otherwise it's warn.
Made-up example of the detail line:
Full exec trust is enabled for: main.
OpenClaw's own suggested fix is:
“Prefer tools.exec.mode="ask" or "allowlist", and reserve "full" for tightly scoped break-glass agents only.”
Why do people see it on a single-user setup?
OpenClaw's exec docs say "no-approval host exec is the default for gateway and node (mode=full)". The audit docs add that security="full" alone:
“is a broad posture warning, not proof of a bug - it is the chosen default for trusted-operator setups; tighten it only when your threat model needs approval or allowlist guardrails.”
OpenClaw's docs describe full as the default, but in 2026.10.5's source this finding only fires when full is written explicitly in your config, in tools.exec.security, tools.exec.mode or an agent's own exec settings. If the setting is missing, the audit treats it as deny for this check. So the finding means someone, or a setup step, chose full in the config. On a single-user setup that's often deliberate, and a warn-level finding is expected.
Related findings
The audit checks other exec settings in the same family. These have their own checkIds:
security.exposure.open_channels_with_exec, when open rooms can reach an agent with exec ontools.exec.auto_allow_skills_enabled, when skill binaries are trusted automaticallytools.exec.allowlist_interpreter_without_strict_inline_eval, when interpreters are allowlisted withoutstrictInlineEval
See every audit finding, explained.
Could it be a false alarm?
This one isn't usually a false alarm. It reports a setting. Check what's in effect for each agent:
openclaw exec-policy show openclaw config get tools.exec
If you've decided full is right for a single-user setup, our guide to audit false alarms explains how to suppress a warn-level finding properly, with a reason written down. Don't suppress the critical version. Fix the open channel instead.
What our free check covers
The free check at /check is paste-only. We check the text you paste: your openclaw security audit output. The check never connects to your Gateway, so it can't read your exec approvals file or see which commands your agent has run. Nothing is installed into the agent you're checking. Green means no issues found in the checks we ran.
The free check puts the most important finding first, in plain English. The full report explains every finding and lists the fixes in order. The check doesn't detect malicious skills. We don't review the code inside skills or MCP servers ourselves, beyond what OpenClaw's own audit reports. The check is free, and a full report is $29. See Limits of the check.
Frequently asked questions
What does tools.exec.security_full_configured mean?
Exec security is set to full for your default agent or a named agent, so host commands run without approval prompts. OpenClaw's audit lists the agents it applies to. It rates the finding warn, or critical when it also finds open DM or group access.
Is security=full a bug?
No. OpenClaw's audit docs say security="full" alone "is a broad posture warning, not proof of a bug - it is the chosen default for trusted-operator setups". They say to tighten it "only when your threat model needs approval or allowlist guardrails".
How do I fix tools.exec.security_full_configured?
Stop the Gateway, run openclaw exec-policy preset cautious, start the Gateway and re-run openclaw security audit --deep. The cautious preset uses an allowlist and asks a human for anything not on it. If the finding is critical, close the open DM or group policy first.
Can your free check see my exec approvals?
No. We check the text you paste, which is where this finding comes from. The check never connects to your Gateway, so it can't read your exec approvals file or see which commands your agent has run.
Sources
- OpenClaw docs: security audit checks (checkId catalog)
- OpenClaw docs: running the security audit (posture warning note)
- OpenClaw docs: exec tool (modes and defaults)
- OpenClaw docs: exec approvals (presets)
- OpenClaw docs: openclaw approvals and openclaw exec-policy (CLI)
- OpenClaw docs: pairing (DM access approval)
- OpenClaw source v2026.10.5: src/security/audit.ts (where the check is defined)
Get the plain-English agent security brief
Agents & AI Security, explained: a short, plain-English email when we have something useful. Free. Unsubscribe any time.
We'll only use your email to send the newsletter. The newsletter is sent by beehiiv, which records opens and link clicks. Privacy policy.