Triage static analysis findings for false positives
An agent reviews each SAST finding in context, marks likely false positives with a reason, and ranks the rest for a security owner.
The use case
Static analysis tools are noisy, and AppSec teams spend hours deciding which findings are real. Semgrep's AI features read the flagged code and its surroundings, label findings as likely true or false positives with a written reason, and suggest fixes for the real ones. The reasoning can be shown to developers so they understand why something matters. A security engineer still spot-checks the false-positive calls, especially for high-severity rules, and decides what gets suppressed. Over time, teams feed back their decisions so the triage matches their codebase.
Example workflow
- Connect the scanner to your repositories
- Agent reviews new findings and labels each with a reason
- Security owner samples false-positive labels each week
- Confirmed issues go to developers with the explanation
- Record decisions so future triage improves
Tools
- Semgrep
Risks and guardrails
- A wrong false-positive label hides a real vulnerability; audit samples
- Do not let the agent suppress critical rules on its own
- Keep an audit trail of who accepted each suppression
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Semgrep docs: AI features overview: docs.semgrep.dev/semgrep-multimodal/overview