How to Show Clients Your OpenClaw Security Work: The Evidence to Keep
Published 2026-10-20 · Updated 2026-10-30 · By The Is My Agent Safe team
Short answer
Give the client dated records, not general reassurances. Keep a dated openclaw security audit --deep --json file, a written reason for every accepted finding, the OpenClaw version and update status, the result of openclaw secrets audit --check, and a verified backup. Each comes from OpenClaw's own commands. Together they show what you checked and when, not that nothing can go wrong.
- Audit: a dated
openclaw security audit --deep --jsonfile, with its summary and findings by checkId and severity. - Accepted findings: suppressions with a written reason. The JSON keeps them under suppressedFindings, and the audit flags that it was filtered.
- Version:
openclaw --versionandopenclaw update status, against the current stable release on OpenClaw's releases page and published advisories. - Secrets and recovery:
openclaw secrets audit --checkexit codes, and a backup checked withopenclaw backup create --verify.
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).
Who this is for
Agencies and consultants who set up or look after OpenClaw for clients, and need to show what they actually did. This page is about the records to keep. For the routine that produces them, see Weekly OpenClaw security checks for clients. For the one-off handoff, see How to audit a client's OpenClaw.
"We hardened it" is a claim. A dated audit file with a reason next to every accepted finding is a record the client can check. Every item below comes from OpenClaw's own commands.
Why clients may ask for records
Some clients will want proof for their own reasons, such as a board, an insurer or a customer questionnaire. If you handle client data, check your contract and any data-protection rules that apply; this guide isn't legal advice.
1. A dated deep audit, as JSON
openclaw security audit --deep --json > "audit-$(date +%Y%m%d).json"
OpenClaw's CLI docs show how to read it:
jq '.summary' audit-YYYYMMDD.json jq '.findings[] | select(.severity=="critical") | .checkId' audit-YYYYMMDD.json
Each finding has a structured checkId, such as gateway.bind_no_auth, and a severity. The audit checks catalog lists every checkId. Keep one file per client per check, so you can show what changed between months.
--deep matters. The CLI docs say it adds best-effort live Gateway checks and plugin-owned security checks that the plain audit doesn't run.
2. A written reason for every accepted finding
Some findings are deliberate choices, such as a plugin the client relies on. OpenClaw lets you accept them in config with security.audit.suppressions. Each suppression matches an exact checkId, can be narrowed by title or detail text, and takes a reason.
This is good evidence, because OpenClaw doesn't hide what you accepted. Per the CLI docs:
- Suppressed findings leave the active
summaryandfindings, but the JSON keeps them undersuppressedFindings. - An info finding,
security.audit.suppressions.active, can't be suppressed and shows readers the audit was filtered.
Agree each accepted finding with the client and write the client's name next to the reason. OpenClaw's running the audit page gives a fix order to follow for everything else, starting with anything open plus tools enabled.
If you used openclaw security audit --fix, keep its JSON as well. The docs say --fix --json output includes both the fix actions and the final report. Note what --fix doesn't do: it doesn't rotate keys, disable tools, change bind or exposure, or remove skills and plugins.
3. Version and update status
openclaw --version openclaw update status
Record the version on every report. OpenClaw's update docs say that when no update is active, openclaw update status labels the saved outcome Last recorded update, with its start time.
Compare the version with the current stable release on OpenClaw's releases page on the day you record it, not a number you remember. Versions below 2026.8.1 are missing fixes for advisories rated high. Write down when you last reviewed OpenClaw's security advisories, and either "none apply" or the advisory IDs you acted on.
4. No plaintext secrets left in config
openclaw secrets audit --check
OpenClaw's secrets CLI docs say audit --check returns exit code 1 on findings, and unresolved references return exit code 2. Record the exit code and date, and clear any findings before you share the record. Never paste the secrets themselves into a report.
5. A backup you've verified
openclaw backup create --verify --output <private-path>
OpenClaw's backup docs say --verify validates the archive straight after writing it, and openclaw backup verify <archive> checks an existing one. Record the date and the verification result.
Don't send the archive as proof. The same docs say archives cover state, config, auth profiles, channel and provider credentials and sessions.
Put it in one evidence table
One short table per client per period is enough. The rows below are a made-up example, not a real client:
| Record | Made-up example |
|---|---|
| Deep audit | audit-20261020.json: 0 critical, 2 warn, 3 info |
| Accepted findings | 1 suppression, reason "client's CRM plugin", agreed with client |
| Version | 2026.10.5. Last recorded update Oct 19 |
| Advisories reviewed | Oct 20: none apply |
| Secrets audit | openclaw secrets audit --check: exit code 0 |
| Backup | Created and verified Oct 20, stored in the client's vault |
| Open items | 2 warnings, owner and date for each |
What the records can't show
The records show what you checked and when. They don't show that an install can't be attacked. The audit's live Gateway checks are best-effort, and a result is only as current as the day you ran it. Say that plainly to the client, and keep the dates on everything.
What our free check covers
The free check at /check is paste-only. Paste a client's openclaw security audit --deep output and openclaw --version. It explains each finding in plain English, checks for risky settings such as API keys handed to skills, and checks the version against published security advisories. Nothing is installed into the agent you're checking. Green means no issues found in the checks we ran. See Limits of the check.
It doesn't run openclaw secrets audit or check backups, doesn't detect malicious skills, and doesn't review MCP servers or skill code. The check is free, and a full report is $29.
You can buy a founding seat on /agencies. $99/month, price locked for your first 12 paid months. After that, it moves to the then-current Agency price, with at least 30 days' email notice. Cancel any time. Offer conditions.
Where next
- OpenClaw: openclaw security (CLI), Security audit checks, openclaw backup
- Weekly OpenClaw security checks for clients
- How to audit a client's OpenClaw
- openclaw security audit explained
Frequently asked questions
What is the single most useful record to keep?
A dated openclaw security audit --deep --json file. OpenClaw's CLI docs show jq '.summary' for the counts and a filter on .findings[] by severity and checkId, so you can compare one month with the next.
How do I show a client that a finding was accepted on purpose?
Use security.audit.suppressions with a reason. OpenClaw's CLI docs say suppressed findings leave the active summary but stay in the JSON under suppressedFindings, and an info finding, security.audit.suppressions.active, shows the audit was filtered.
Does a clean audit prove the install can't be attacked?
No. It shows what OpenClaw's audit found at the time you ran it. OpenClaw's own docs call the --deep live Gateway checks best-effort, and new releases and advisories keep coming. Present it as a dated result, not a promise.
Should I send the client the backup archive as proof?
No. OpenClaw's backup docs say archives cover state, config, auth profiles, channel and provider credentials and sessions. Send the verification result and the date, and keep the archive itself where only you and the client can reach it.
Can Is My Agent Safe produce this evidence for me?
Partly. Paste an openclaw security audit --deep output and openclaw --version into the free check at /check for a plain-English explanation of the findings, a check for risky settings, and a version check against published security advisories. It doesn't detect malicious skills, review MCP servers or skill code, run openclaw secrets audit or check backups.
Sources
- OpenClaw docs: openclaw security (CLI: JSON output, suppressions, --fix)
- OpenClaw docs: security audit checks
- OpenClaw docs: running the security audit
- OpenClaw docs: openclaw update (CLI)
- OpenClaw docs: openclaw secrets (CLI)
- OpenClaw docs: openclaw backup (CLI)
- OpenClaw GitHub security advisories
- OpenClaw releases (GitHub)
- OpenClaw release v2026.8.1 (Aug 31, 2026)
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.