#048Guide3 min

Handle several vulnerabilities without losing track

Use correction search, status filters, and finding evidence to keep parallel repairs understandable.

Hackaiz team · updated 21/11/2026

Illustration: two developers review a change before approving it.
Illustration: two developers review a change before approving it. Illustrative photo, not a screenshot.

01Separate findings from correction work

A list of vulnerabilities and a list of corrections answer different questions. A finding describes a security problem and its evidence; a correction describes work proposed to address it. Start by reading the findings you intend to handle and identifying their affected files or components. Two findings in one file may need different protections, while one shared authorization helper may explain symptoms in several places. Do not equate the number of rows with the number of independent engineering tasks.

Make a short worklist in your team’s normal tracking system that links each intended change to its finding. This is a recommended practice, not a claim that Hackaiz includes a ticket assignment system. Keep separate entries when the required permissions, reviewers, or deployment steps differ. For instance, replacing an exposed provider key and tightening a database policy should not disappear behind a single vague task called “security fixes.” Their completion evidence is different.

02Find the next correction to handle

Open Fixes and use Search corrections. Its placeholder, Finding, file, platform…, shows useful starting points: a file path, the finding wording, or the affected platform. Combine the search with Correction status and Platform rather than repeatedly scanning the full list. The displayed count shows how many corrections remain visible out of the total. If the list unexpectedly becomes empty, use Reset filters before concluding that a correction has vanished.

Use the quick filters deliberately. Ready to review gathers work awaiting a decision; Applied identifies applied changes; Verified isolates completed verification. The full status list also distinguishes Requested, Preparing, Ready for review, Applied · verify, Verified, and Failed. Those stages help you decide whether the next task is to wait for preparation, review a change, investigate a failed attempt, or finish verification. Searching and filtering reorganize the view; they do not launch additional tests.

03Review risk and scope before grouping changes

Customer-facing finding levels are Danger, To fix, and To watch. Use them to orient your review, then inspect what the evidence actually establishes and which assets are affected. A publicly exposed credential and a configuration observation do not call for the same immediate response. Avoid silently downgrading a serious problem because its correction is inconvenient. If several proposed repairs touch the same function, have a developer check how they interact before merging them in succession.

For each ready proposal, inspect Change details and validation and follow its finding. Review and approve preview is distinct from Review in GitHub when a pull request link exists. Hackaiz publishes approved proposals as draft GitHub pull requests for human review; it never merges them automatically. Keep a recognizable purpose for each review. Grouping changes for your own convenience should not erase which finding a patch addresses or whether a particular fix was included in the deployed revision.

04Close each item at the right stage

Do not count every applied correction as a finished security outcome. A repaired scenario needs replay evidence, passing automated tests, and correspondence between the corrected commit and deployment to be classified as verified. When one change is verified and another remains applied but unverified, report that distinction. An inconclusive replay or unavailable tests are reasons to keep an item open, not reasons to silently move it into the completed group.

Analyses and correction operations consume credits, so select the work you actually want to perform. Unticked checks are neither run nor billed. Before any website or API analysis, domain ownership proof is required through _hackaiz DNS TXT, the HTTPS verification file, or the project AI creating that file. If your worklist covers multiple environments, keep each result attached to the address and source version tested. This prevents evidence from a staging repair being used to close a production issue by mistake.

Frequently asked questions

Was this article helpful?

Related articles