#072Guide4 min
Run a security check before deployment
Choose between checking local code and analyzing a connected platform, then review coverage before releasing.
Hackaiz team · updated 21/11/2026

01Decide what your release check must cover
A pre-deployment check is useful only if you know which code or system it examined. Hackaiz offers an editor scan of the open workspace and analyses of platforms already saved in your account. These are different operations. A workspace scan examines eligible files that you confirm sending from your computer. A platform analysis examines the saved target and its connected sources. Neither operation should be described as a test of your next deployment merely because you launched it just before publishing.
For example, reviewing an API handler in your editor can highlight a risky database query, while analyzing a deployed website concerns the deployed target. If your release changes authentication, identify both the changed source files and the environment where the change will run. Keep the release decision separate from the existence of a recent report: a report may concern an older deployment or a different platform.
02Check the code you are about to ship
In the Hackaiz editor extension, use “Scan current file” for a focused change or “Scan this workspace” for the eligible files in the open folder. Connect a personal key first; launching checks requires the launch permission. Review the files and exclusions before confirming the scan. These checks use account credits, and the result shows the actual charge. The extension does not edit your files, commit a change or publish a release.
Exclusions matter to your decision. Dependency folders, generated output, secret files such as .env and oversized files are filtered out. A result with no findings therefore does not mean every file in the repository was inspected. If the run is incomplete, narrow the folder or inspect a relevant file separately. Each additional scan is a separate credit-using operation, not a continuation that silently fills all missing coverage.
- 1Open the folder containing the release source.
- 2Run the workspace or current-file command and review what will be sent.
- 3Read findings together with the completion and coverage information.
03Use a connected run without confusing it with a commit test
The prepared command-line client can request an analysis with “hackaiz analyze PLATFORM_UUID STABLE_REQUEST_KEY”. Its verify command checks the report tied to a specific request, not whichever report happens to be latest. The CLI source exists in tools/hackaiz-cli; the integration page says it is not published on npm and autonomous processing must be deployed and synchronized. Do not assume a public package installation or a working production runner merely from the presence of source.
For a website or API, prove domain control before analysis. Use the DNS TXT record named _hackaiz, publish /.well-known/hackaiz-verification.txt over HTTPS, or ask the AI editing your project to create that verification file using the instruction from Connect. Repository access does not replace domain proof. In the web analysis settings, unchecked checks are neither run nor billed; record those omissions when explaining what the release check covered.
04Review the result before approving release
Use the customer-facing groups Danger, To fix and To watch to organize review. Open important findings and compare the cited file or target with the release you intend to deploy. A saved-platform verification returns an inconclusive outcome for incomplete or unavailable analysis. Even a passing result is not a certification and does not establish that the saved platform matches the proposed commit. Your team must decide which findings and missing checks block publication.
If a repository finding leads to a Hackaiz correction, review the proposed change and its tests before approval. Fixes are draft GitHub pull requests for human review; they are never merged automatically. After applying a change, check the affected behavior and obtain a relevant new analysis when needed. Do not use a clean report from before the fix as evidence that the correction works, and do not treat a successful analysis launch as a successful security gate.
Frequently asked questions
Take the next step
Was this article helpful?