Standing service github-actions-workflows-kept-hardened-each-month · revised 12 October 2026
Standing service
Keep your GitHub Actions pinned, minimal and agent-ready, month after month
Each month your workflows are re-scanned for new unpinned actions, wider permissions and untrusted text in prompts, and fixes arrive as pull requests you merge.
This starts a conversation by email. Nothing is charged, and nothing is read or run, until we have agreed scope, access and terms with you in writing.
The responsibility you hand over
Workflows drift. A new job adds an unpinned third-party action, a copied snippet widens token permissions, or an agent step is added without limits. The risk returns with no one looking.
Who it’s for: A team that has hardened its workflows once and uses AI agents or third-party actions in them.
Usually starts when: Workflows change every week, and a new unpinned action or wider permission can undo the work.
The result: Each month every workflow is re-scanned with the same analyser, new findings are fixed in a pull request you merge, pins are updated, and a short summary shows what changed since last month. We hold no secrets and change no settings.
What stays true, and what we do about it
No response-time guarantee is published for this new service. A target for sending each month's result is agreed in writing before it starts.
The service runs once a month. It is not a pager and has no round-the-clock cover.
What must remain true
- Each month every workflow is scanned and any new finding at the agreed severity has a fix pull request or a recorded exception
- Every agent step still holds no secrets except a restricted model key with a spend cap and a minimal-permission workflow token, and no direct write tools
What we watch
- The workflow files in the repository, read-only
- The analyser's output each month
When something happens
Scroll the table sideways to read it all.
| When | What we do |
|---|---|
| A month closes. | We scan, open a fix pull request for new findings and send the summary. |
| A new agent step or new third-party action appears. | We flag it first in the summary and check its permissions and pins. Linked job: Harden Actions before adding AI agents |
| A fix pull request breaks a workflow. | We close it and list the finding as an exception for your decision. |
We do on our own
- Read workflow files and run the analyser
- Open fix pull requests on a branch
We ask you first
- Any exception
- Wider access
- Anything that changes secrets or settings
We escalate to you when
- A workflow shows signs of abuse
- An agent step gains a secret other than its restricted model key or a minimal-permission workflow token, or a direct write tool
- Pull requests stay unmerged two months
How you know it held. Each month the scan output is attached so your maintainer can re-run the analyser on the same commit and get the same findings.
How we keep it true
This service is never finished. Each month's summary shows whether what it watches stayed true, and it continues until you end it.
Tighten your GitHub Actions before you add an AI agent to them Job Optional
Workflows scanned and fixed: actions pinned to SHAs, minimum token permissions, no untrusted text reaching a shell or prompt, and agent steps with a restricted model key and no direct write tools.
Bought separately, once
The one-off job that hardens the workflows first and runs the planted-instruction tests.
What is included, and what is not
- The monthly scan result
- A fix pull request when there are new findings
- A short summary
Included
- One repository on GitHub Actions with up to 25 workflow files
- A monthly scan, with fixes for new findings as a pull request
- A check that agent steps still hold no secrets except a restricted model key with a spend cap and a minimal-permission workflow token, and no direct write tools
- A short summary of changes since last month
Not included
- Guarantees of security
- Review of third-party action source code
- Organisation settings or secret rotation
- Adding agent workflows
How we know it’s done
Agreed with you before work starts. Each check produces evidence you keep.
Each month the scan output is attached and any new finding at the agreed severity has a fix pull request or a recorded exception.
Evidence: The scan output and the pull request list.
Every agent step holds no secrets except a restricted model key with a spend cap and a minimal-permission workflow token, and no direct write tools, shown by a permission listing.
Evidence: The monthly permission listing.
Your maintainer can re-run the analyser on the same commit and get the same findings.
Evidence: The commit identifier and the analyser version.
Sign-off. You read each summary and tell us whether the severity and exceptions still suit you. A month counts as delivered when you have received the result and the summary.
If it fails. If we cannot read the workflows we say so and do not guess. If the service is not working for you, you can end it at the end of any month.
When it fits, and when we stop
It fits when
- Workflows that were hardened once, by us or by you
- A maintainer who merges pull requests
- Read-only access to the workflow files
We stop and tell you if
- A workflow needs production secrets to test
- Two months pass with pull requests unmerged
- Signs that a workflow has been abused
What could go wrong
Fix pull requests can be closed or reverted. Ending the service leaves your workflows as they are.
Scroll the table sideways to read it all.
| Risk | How we handle it |
|---|---|
| A fix breaks CI. | Changed workflows run on the branch first, and a break turns into an exception for your decision. |
| A clean scan is read as security. | Every summary says it is not a guarantee. This is evidence for your engineers, not assurance, certification or an audit opinion. |
| Pull requests pile up unmerged. | The service pauses after two months without a merge. |
A separate automated reviewer checks the work before hand-over. Your maintainer reviews and merges every pull request. No human supervisor is included unless your agreement names one, and at launch the work is largely automated.
Questions
Do I need the one-off hardening first?
Yes, or an equivalent you did yourself. The monthly service starts from a hardened baseline; the one-off job is bought separately.
Do you hold our secrets?
No. We hold none and change no settings.
Is a clean scan a guarantee?
No. It only says the analyser found nothing at the agreed severity.
Also part of
Send an enquiry
Send us
- How many workflow files you have and whether any use AI agents
- Who merges
- No workflow code, secrets or logins in the first enquiry
Later, once you agree
- Read-only access to workflow files through a company-controlled route
- A branch for pull requests
- The name of your maintainer
You keep the repository, secrets and merge rights. We work through a dedicated bot account with the least access that works (read-only first), work on a fork or branch, and you merge, hold no secrets and change no settings. Our automated agents read the files you send in a sandbox; we make no "no training" or "no storage" promise here, and the agreement names the model provider, country, retention and training status before anything is shared.
Email fallback: open your mail app
If website submission is unavailable, review and send the fallback email yourself. An email fallback is not a website receipt. Or write to hello@syntheticindustry.ai with “github-actions-workflows-kept-hardened-each-month” as the subject.