OpenClaw plugins.code_safety Finding: Real Risk or False Alarm?

Published 2026-10-20 · Updated 2026-10-20 · By The Is My Agent Safe team

Short answer

OpenClaw's deep audit found suspicious or dangerous patterns in an installed plugin's code. Find which plugin and file it names first. If you didn't knowingly install that plugin, or can't explain the flagged code, disable it. An open GitHub report (#82469) describes critical findings on bundled plugins' test files; it isn't confirmed.

  • plugins.code_safety only appears with openclaw security audit --deep. It can be warn or critical.
  • First, find the plugin, file and rule in the finding's title and detail. Then decide whether you trust that plugin.
  • If you don't, disable it with openclaw plugins disable <id> and re-run the audit.
  • Issue #82469 is an open report of false alarms on bundled plugins' test files. OpenClaw's maintainers haven't confirmed it or shipped a fix.

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 which plugin and file the finding is about:

openclaw security audit --deep --json | jq '.findings[] | select(.checkId|startswith("plugins.code_safety"))'

The title names the plugin, and the detail lists the rule that matched (such as env-harvesting) and the file and line. Then ask one question: do you trust this plugin?

```bash

openclaw plugins disable <id>

openclaw security audit --deep

```

If you don't need it, openclaw plugins uninstall <id> removes it. The command also takes --dry-run. If you think the plugin has already run harmful code, treat the keys your agent can reach as exposed. Our guide to a leaked API key covers what to do in the first hour.

OpenClaw's plugin docs say to "treat plugin installs like running code". Its audit triage order ends with "Plugins: load only what you explicitly trust."

What does plugins.code_safety mean?

OpenClaw's security audit checks page lists plugins.code_safety as warn or critical: "Plugin code scan found suspicious or dangerous patterns (--deep only)". A plain openclaw security audit doesn't run this check, so you only see it with --deep.

The same page lists related checkIds that are easy to confuse with it:

checkIdSeverityWhat it means (from OpenClaw's docs)
plugins.code_safetywarn/criticalPlugin code scan found suspicious or dangerous patterns
plugins.code_safety.entry_pathwarnPlugin entry path points into hidden or node_modules locations
plugins.code_safety.entry_escapecriticalPlugin entry escapes the plugin directory
plugins.code_safety.manifest_parse_errorwarnPlugin manifest could not be parsed during the scan
plugins.code_safety.scan_failedwarnPlugin code scan could not complete
plugins.code_safety.scan_truncatedwarnPlugin code scan reached its file or traversal budget

The code check has limits. The docs say it checks up to 500 source files per plugin, skips files larger than 1 MiB, and stops after 100,000 directory entries. A scan_truncated warning "means some files were not checked; it does not identify dangerous code."

Could it be a false alarm?

There's one open report about this checkId. Issue #82469, filed in May 2026 against OpenClaw 2026.4.29, says the check raised critical env-harvesting findings on test files (such as *.test.ts) inside bundled plugins. The reporter says those files don't load when the plugin runs, and that the rule matched the word fetch ( inside a test's description text. As of October 20, 2026, the issue is open and labeled for maintainer and security review. OpenClaw's maintainers haven't confirmed it or shipped a fix.

Your finding may match that report if all of these are true:

If it doesn't match all three, treat the finding as real until you've read the flagged code.

If you've confirmed a finding doesn't apply, OpenClaw's security CLI docs describe suppressions: each one "matches an exact checkId", can be narrowed with titleIncludes or detailIncludes. The docs' example also gives a reason, so write down why. Suppressed findings stay in the JSON output under suppressedFindings. Our guide to audit false alarms walks through it, and every audit finding, explained covers the other plugins.* checks.

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. We don't review the code inside plugins, skills or MCP servers ourselves, beyond what OpenClaw's own audit reports, and we don't detect malicious plugins or skills. 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 is free, and a full report is $29. See Limits of the check.

Frequently asked questions

What does plugins.code_safety mean?

OpenClaw's security audit, run with --deep, checked an installed plugin's code and found suspicious or dangerous patterns. OpenClaw lists it as warn or critical, depending on what was found.

Is plugins.code_safety a false positive?

Sometimes it may be. Issue #82469 on OpenClaw's GitHub reports critical env-harvesting findings on test files inside bundled plugins. It's an open report. OpenClaw's maintainers haven't confirmed it or shipped a fix. Check the file and rule your finding names before you decide.

What's the difference between plugins.code_safety and plugins.code_safety.scan_truncated?

plugins.code_safety means the code check found a suspicious pattern. scan_truncated means the check hit its file or traversal limit, so some files weren't checked. OpenClaw's docs say a scan_truncated warning "does not identify dangerous code".

Does your free check review my plugin code?

No. We check the text you paste, which is where OpenClaw's plugins.code_safety finding comes from. We don't review the code inside plugins or skills ourselves, beyond what OpenClaw's own audit reports, and we don't detect malicious plugins or skills.

Sources

  1. OpenClaw docs: security audit checks (checkId catalog)
  2. OpenClaw docs: openclaw security (CLI), including suppressions
  3. OpenClaw docs: running the security audit
  4. OpenClaw docs: plugins
  5. OpenClaw docs: openclaw plugins (CLI)
  6. GitHub issue openclaw/openclaw#82469: plugins.code_safety false positives on bundled plugin test files (open)