What problem does it solve? Security teams accumulate noisy backlogs of findings from scanners, CVEs, advisories, bug bounty reports, and Jira/Linear tickets, and manually verifying whether each claim actually affects the current codebase is slow and inconsistent. ## Core Features & Use Cases - Static Evidence-Based Verdicts: Assigns one of three verdicts (confirmed, not_actionable, needs_review) per finding using source-to-sink tracing, reachability analysis, and SECURITY.md policy gates, without running any code. - Multi-Source Intake: Imports findings from GitHub code scanning, Dependabot alerts, security advisories, Jira, and Linear via REST or MCP connectors, normalizing them into a common triage record. - Exploitability Stack Ranking: Ranks confirmed and needs_review findings in separate priority queues and emits a machine-readable triage-finding/v0 JSON contract. - Use Case: A team imports 40 open Dependabot and code-scanning alerts from GitHub and receives a ranked, evidence-backed action queue showing which vulnerabilities are actually reachable in their shipped code. ## Quick Start Triage the open security findings from this GitHub repository and tell me which ones are confirmed exploitable in the current code.