#084Guide4 min

Integrate Hackaiz checks into CI/CD

Choose between checking submitted components and analyzing a saved platform, then protect credentials and interpret the result.

Hackaiz team · updated 21/11/2026

Illustration: a developer uses an editor and a terminal.
Illustration: a developer uses an editor and a terminal. Illustrative photo, not a screenshot.

01Choose what your pipeline is checking

Continuous integration and delivery, often shortened to CI/CD, runs repeatable tasks when code changes and prepares releases. There are distinct Hackaiz integration paths. The official import-gate action checks submitted components. The connected workflow requests analysis of a platform already saved in Hackaiz. Do not treat their outputs as interchangeable: a configuration check is not a complete application analysis, and a saved-platform analysis does not necessarily examine the pull-request head commit.

Start from Integrations and choose a path for your actual question. For a component or workflow review, the documented action is Hacktualiz/import-gate@v1. For a saved authorized target, the template is https://hackaiz.com/hackaiz-connected-action.yml and uses the prepared tools/hackaiz-cli client. The connected client is not published on npm and needs the autonomous processing service configured and deployed. A supplied template is not evidence of a successful run in your repository.

02Configure the official component action carefully

The Integrations page documents adding HACKAIZ_API_KEY in GitHub under Settings → Secrets and variables → Actions, then creating .github/workflows/hackaiz.yml with uses: Hacktualiz/import-gate@v1 and the api-key input set to ${{ secrets.HACKAIZ_API_KEY }}. GitHub’s Choose a workflow screen is not an action catalogue: write and review the workflow file in the repository. Check the action’s own documented inputs before deciding which content it will submit.

This path performs a static check of submitted text; it does not execute the component, observe runtime behaviour or read all installed dependency code. For example, checking a package manifest or a workflow can inform a review of a proposed addition without proving the whole application secure. The documented default makes Danger or an incomplete result fail the job. Each run uses credits. Optional SARIF output, a structured findings format, can feed GitHub’s Security tab where repository permissions and Code scanning options allow it.

03Understand the connected template before enabling it

The connected template runs for pull requests from the same repository and excludes forks. It checks out the default branch, uses contents: read, and grants pull-requests: write to comment a result link. It supplies the API key as a secret and the project identifier through HACKAIZ_PROJECT_ID in repository variables. Its request key includes the repository, pull request and head commit identifiers to identify the request; that key does not prove the commit was analyzed.

The analysis target is still the saved platform. For any website or API, prove ownership first with DNS TXT _hackaiz or the HTTPS file /.well-known/hackaiz-verification.txt, which your project’s AI can help create. Keep consent current. The client submits a request and polls its status, while autonomous work continues if the client closes. Preserve the returned request identifier and query that same request after a lost connection rather than creating another analysis. The template avoids cancelling an in-progress job when another event arrives.

04Protect the pipeline and review its evidence

Never expose HACKAIZ_API_KEY to untrusted fork code, print it in job logs or paste it into a pull-request comment. Review workflow changes like application code: permissions and event triggers determine who can run code with your credentials. Retain the project reference, request identifier, job outcome and actual analysis status. A green launch job is not always complete evidence because the connected analyze command also accepts partial completion. Review coverage before using its output for a release decision.

The connected template comments a link; it neither modifies source nor merges code and does not by itself create a deployment gate. If your release requires a stricter decision, add a reviewed verification stage and explicit dependencies in your delivery system. Analysis work consumes credits; unticked checks are not run or billed. Hackaiz corrections remain draft GitHub pull requests for human review and approval, never automatic merges. Keep security review, change approval and release authorization as separate, visible decisions.

Frequently asked questions

Was this article helpful?

Related articles