OpenClaw gateway.probe_failed and Other "Audit Didn't Finish" Findings
Published 2026-10-20 · Updated 2026-10-20 · By The Is My Agent Safe team
Short answer
openclaw security audit --deep tried to check your running Gateway and couldn't, so part of the audit didn't run. It's not a sign of an attack. Fix it first: make sure the Gateway is running (openclaw gateway status), run openclaw status --all to see the connection or auth error, fix that, and run openclaw security audit --deep again.
- OpenClaw rates gateway.probe_failed warn: "Live Gateway check failed (--deep only)". The detail is the error the check got.
- Most often the Gateway isn't running, or the audit couldn't sign in to it. Set OPENCLAW_GATEWAY_TOKEN (or OPENCLAW_GATEWAY_PASSWORD) in the same shell and try again.
- A "missing scope: operator.read" error on a loopback Gateway was a bug. The fix for one cause (#159121) is in 2026.10.5. A related report (#128217) is still open.
- Our check treats this finding, and the others on this page, as an audit that didn't finish, so it won't show a green result.
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?
Find out why the live check failed:
openclaw gateway status openclaw status --all
Then fix what they show:
- The Gateway isn't running. Start it with
openclaw gateway start. The deep check needs a running Gateway. - An auth error. The audit couldn't sign in. Set the Gateway token or password in the same shell, for example
export OPENCLAW_GATEWAY_TOKEN=…, or pass it for one run with--tokenor--password. OpenClaw's security CLI docs say these "only override deep-check auth for that command invocation; they do not rewrite config or SecretRef mappings". missing scope: operator.read. See the false-alarm section below. Update to 2026.10.5 first.
Then run the audit again:
openclaw security audit --deep
When it comes back without gateway.probe_failed, the live check ran.
What does gateway.probe_failed mean?
OpenClaw's security audit checks page lists it as warn: "Live Gateway check failed (--deep only)". The fix it points to is "gateway reachability/auth".
In OpenClaw 2026.10.5's source, the finding appears when --deep tried the live Gateway check and it didn't succeed. The detail is the error the check got, or "gateway unreachable" if there wasn't one. Made-up example of the detail:
connect ECONNREFUSED 127.0.0.1:18789
OpenClaw's own suggested fix is:
“Run "openclaw status --all" to debug connectivity/auth, then re-run "openclaw security audit --deep".”
It's not a sign of an attack. It means the audit's live check of your running Gateway didn't happen, so you don't have those results.
Other findings that mean the audit didn't finish
Our check treats these OpenClaw findings the same way: part of the audit didn't run, or its results may be filtered. Descriptions are from OpenClaw's checks table.
| checkId | OpenClaw's description | Where to look |
|---|---|---|
gateway.probe_auth_secretref_unavailable | "Deep check could not resolve auth SecretRefs in this command path" | Set OPENCLAW_GATEWAY_TOKEN or OPENCLAW_GATEWAY_PASSWORD in this shell, or resolve the secret provider |
plugins.code_safety.scan_failed | "Plugin code scan could not complete (--deep only)" | Plugin path and scan environment |
skills.code_safety.scan_failed | "Skill code scan could not complete (--deep only)" | Skill scan environment |
skills.workspace.scan_truncated | "Workspace skill scan reached a traversal limit or could not read part of the tree" | Directory access and the size of the workspace skills/ tree |
plugins.<pluginId>.security_audit_failed | "A plugin-owned security audit collector threw an error" | That plugin's audit collector |
channels.<provider>.account.read_only_resolution | "A channel account could not be fully resolved for audit (missing secret/gateway)" | "check that referenced secrets can be resolved, or run against a live gateway snapshot" |
sandbox.browser_container.docker_probe_timeout | "Docker label check for the browser container timed out" | Docker daemon reachability |
security.audit.suppressions.active | "Audit output has configured suppressions and may be filtered" | security.audit.suppressions in your config |
The same page gives the limits of OpenClaw's deep code scans: up to 500 source files per plugin or skill, skipping files larger than 1 MiB, and stopping after 100,000 directory entries. Workspace skill scans stop at 2,000 skill files and 40,000 directory visits.
plugins.code_safety.scan_truncated and skills.code_safety.scan_truncated aren't in this group. OpenClaw's docs say they mean "some files were not checked; it does not identify dangerous code". Our check shows them as ordinary findings. See plugins.code_safety findings.
Could it be a false alarm?
It can be. The error missing scope: operator.read on a local Gateway has been reported more than once:
- Issue #46836, filed March 2026 on 2026.3.13, was closed as completed later that month.
- Issue #159121, filed September 26, 2026, said the deep check on a loopback Gateway looked up the cached device token in the wrong place and connected without its identity. It was closed by pull request #159169, merged September 28, 2026. GitHub's comparison shows that change is included in 2026.10.5.
- Issue #128217, filed August 23, 2026, is still open. An automated reply in August said the fix was already on the main branch. A comment on October 10, 2026 says it still happened on 2026.9.8 and 2026.9.9 in setups with several Gateways on one host, each under its own user.
If you're on 2026.10.5 and still get missing scope: operator.read while openclaw status --all shows the Gateway reachable and signed in, add your details to #128217 rather than opening a new report. Our guide to audit false alarms covers what to include. Don't suppress this finding: while it's there, part of the audit isn't running.
What our free check covers
The free check at /check is paste-only. We check the text you paste: your openclaw security audit --deep output. The check never connects to your Gateway, so it can't run the live check for you. If your paste includes gateway.probe_failed or another finding in the table above, our check marks the audit as not finished and won't show a green result. We use our own scale: an OpenClaw critical finding still shows red, and most OpenClaw warn findings, including gateway.probe_failed itself, show amber. Otherwise the result is grey. 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 gateway.probe_failed mean?
During openclaw security audit --deep, OpenClaw tried to connect to your running Gateway and couldn't. The finding's detail is the error it got, or "gateway unreachable". OpenClaw rates it warn.
How do I fix gateway.probe_failed?
Check the Gateway is running with openclaw gateway status, then run openclaw status --all to see the connection or auth error. If it's an auth error, set OPENCLAW_GATEWAY_TOKEN or OPENCLAW_GATEWAY_PASSWORD in the same shell. Then re-run openclaw security audit --deep.
Why do I get "missing scope: operator.read" on a local Gateway?
At least one cause was a bug in how the deep check signed in. Issue #159121 was closed by a fix that is in 2026.10.5, so update first. Issue #128217, about the same error, is still open, with a report from October 2026 that it still happened on 2026.9.8 and 2026.9.9 in setups with several Gateways on one host.
What does gateway.probe_auth_secretref_unavailable mean?
The Gateway's token or password is stored as a secret reference that the audit couldn't resolve in this shell. OpenClaw suggests setting OPENCLAW_GATEWAY_TOKEN or OPENCLAW_GATEWAY_PASSWORD in this shell or resolving the secret provider, then re-running openclaw security audit --deep.
Can your free check give me a green result with this finding?
No. We check the text you paste. When it includes gateway.probe_failed or the other findings on this page, our check marks the audit as not finished, so the result can't be green. We use our own scale: an OpenClaw critical finding still shows red, and most OpenClaw warn findings, including gateway.probe_failed itself, show amber. Otherwise the result is grey.
Sources
- OpenClaw docs: security audit checks (checkId catalog and deep scan limits)
- OpenClaw docs: openclaw security (--deep, --token, --password, SecretRefs)
- OpenClaw docs: Gateway service (gateway status)
- OpenClaw source v2026.10.5: src/security/audit-deep-probe-findings.ts (where the checks are defined)
- OpenClaw source v2026.10.5: src/security/audit-gateway-probe.ts (the live check)
- GitHub issue #159121 (closed, fixed by pull request #159169): missing scope operator.read on loopback
- GitHub issue #128217 (open): deep audit fails its own gateway probe
- GitHub issue #46836 (closed): missing scope operator.read on loopback
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.