PHANTOM Assess.
Autonomous testing with verifiable proof.
Connect an authorized private repository, define scope, launch an idempotent scan, follow the live execution stream, and inspect evidence-backed findings with commit deltas.
What ships in the product today — zero aspirational claims
Connect authorized private repositories through the official GitHub App with fine-grained, read-only permissions (contents:read, metadata:read). Access is revocable instantly from your GitHub settings; personal access tokens are never used.
Select target branches, tags, or exact commit SHAs resolved through GitHub's Git Data API. Idempotent scan launches enforce deterministic Idempotency-Keys, guaranteeing duplicate clicks cannot trigger redundant scans or double budget reservations.
Follow scan sessions in real time via Server-Sent Events (SSE). Watch stage chips animate sequentially: ORACLE recon → CORTEX hypotheses → HYDRA testing → WEAVER chaining → SCRIBE verified reporting.
Verified findings, code snippets, and remediation diffs are exportable in human-readable Markdown, structured JSON, and standardized OASIS SARIF 2.1.0 for instant ingestion into GitHub Code Scanning and enterprise SIEMs.
Push a remediation commit and trigger an immediate rescan. PHANTOM categorizes findings into Resolved, Persistent, and New, providing an honest security delta between commit lineages.
If upstream provider failure occurs (e.g. model outage), PHANTOM switches to deterministic local fallback rules. The scan session, API response, and all exported reports are prominently marked [DEGRADED] with exact fallback stages, offering a one-click rescan at full fidelity.
Fail-Closed Resource Admission & Spend Caps
Every spend-producing scan evaluates four layers of pre-flight admission before a worker job is admitted.
spend_limit_exceeded or concurrency_limit_exceeded response without creating orphan jobs or phantom spend.DESIGN PARTNER ONBOARDING