AI-Assisted Review
Hashiro's AI assessment engine generates vulnerability findings as candidates, draft findings that require human review before being finalized. The AI review workflow streamlines triaging, validating, and refining automated discoveries.

Screenshots use fictional demonstration data.
Candidate Review Panel
When AI assessments produce findings, they appear in the candidate review panel within the project's Issues tab. Each candidate can be actioned with:
| Action | Effect |
|---|---|
| Accept | Confirms the finding as valid; it becomes a full issue |
| Reject | Dismisses the finding as a false positive or irrelevant |
| Needs Evidence | Requests additional proof before making a decision |
| Merge | Combines with an existing finding covering the same vulnerability |
TIP
Candidates can be marked as publicly visible to clients before full verification. This lets clients see what automated testing discovered while the provider completes the review.
AI Demonstration Regeneration
For accepted or in-progress findings, the AI can regenerate the exploitation demonstration:
- Select a finding and request demonstration regeneration
- The AI produces an updated step-by-step proof of exploitation
- Review the generated content in a preview
- Apply the demonstration to update the finding, or discard it
This is useful when the original demonstration needs clearer steps, better formatting, or additional context.
AI Merge Proposals
When the AI detects multiple findings describing the same or overlapping vulnerabilities, it can propose a merge:
- The AI analyzes similarity between findings
- A merge proposal shows which findings would be combined
- You can accept or reject the merge
- Accepted merges consolidate findings into a single, comprehensive issue
This prevents duplication and produces cleaner reports.
Workflow Integration
AI-assisted review integrates with the standard issue lifecycle:
AI Assessment → Candidate → Review → Accepted (becomes Open issue)
→ Rejected (archived)
→ Merged (combined with existing)All review actions are recorded in the finding's history timeline.