#104Guide4 min
Keep evidence of a security check
Preserve the result, scope and limitations so a later reviewer can understand what ran.
Hackaiz team · updated 21/11/2026

01Identify the check before saving its output
A security report is a dated observation of a particular target, not a permanent statement about an application. Before filing evidence, note the project, analysis reference, target, environment and version where known. Use the project’s analysis history to reopen the result you actually want to preserve. Do not substitute the latest analysis simply because its summary looks cleaner: a later run may examine a different deployment or a different selection of checks.
Keep the authorization context in your own records. Websites and APIs require domain ownership proof before analysis, using the DNS TXT record _hackaiz, the HTTPS file /.well-known/hackaiz-verification.txt, or the project AI creating that file. Never archive a live access token as proof of permission. A statement identifying the authorized asset and responsible person is safer than copying credentials into an evidence folder that other reviewers can read.
- 1Open the project and locate the relevant analysis.
- 2Write down the analysis reference and environment in your evidence index.
- 3Check whether the result is completed, partial or stopped before describing its meaning.
02Choose exports for the question being asked
In a completed combined report, Download in the Findings area saves a Markdown document, a text format readable by common document tools. It includes the target names, export date, target statuses and finding summaries with rules. This export is useful as an index, but do not assume it contains every detail shown on an individual finding page. Preserve the finding references and any necessary supporting evidence separately in your controlled records.
For control-level context, use Download (JSON) in the technical alignment area. JSON is a structured text format that retains the analysis identifier, generation time, standards, control identifiers, states and available references such as rule, file and line. These two exports answer different questions: one summarizes findings, while the other preserves technical coverage states. Keep their original filenames or record any renaming in your evidence index so another reviewer can trace the files back to their source.
03Save what was not established as well
The strongest evidence package includes its limits. Keep statuses such as Partly examined and Not examined by this analysis visible, along with any control marked Must be proven by a person. A user declaration is not independently verified evidence. If an analysis is partial, identify which conclusion remains unsupported instead of presenting the available output as a complete assessment. Unchecked checks are not executed or billed, and their absence should be recorded as a scope limit.
Be careful with dates: an export generation time is not necessarily the execution time of the check. Write both separately when available, rather than silently treating a downloaded file as a fresh analysis. Likewise, a screenshot of a severity total is not enough to show what was tested. Retain the target context, the relevant evidence and customer-facing severity, Danger, To fix or To watch, so the reader can interpret the result without guessing from a dashboard image.
04Protect the archive and link subsequent work
Store the collection in a location your organization controls, with access appropriate to the evidence. Finding details can reveal internal paths, application behavior or sensitive fragments. Review material before sharing it and remove unnecessary secrets from a distribution copy without altering the original restricted record. Define a retention period and a person responsible for the collection in your own procedure; this guide does not promise a built-in retention policy or an immutable archive.
When a correction follows, add the human approval, draft GitHub pull request and deployment reference, then link a later authorized recheck. A proposed fix is not proof of deployment, and Hackaiz never merges automatically. Keep the original report rather than overwriting it with the recheck, so the sequence remains understandable. New analyses and correction preparation use credits. Your archive demonstrates the work and results you retained; it does not certify an application or guarantee that every weakness has been found.
Frequently asked questions
Take the next step
Was this article helpful?