Post-mortems · deep dives

Case studies

Long-form analyses of high-severity findings. Each case study includes the original evidence bundle, per-provider reasoning, and reproducibility instructions.

• highsecurityJune 2026

  n8n-workflows CVE-2025-55526 — api_server.py path traversal

  Neutral-label scan — Opus and GPT-5 both caught the path traversal and graded it at the right severity.

  read case study →
• criticalmethodologyJune 2026

  Zcash Orchard counterfeiting bug

  Re-ran our gate against the 2021 introducing commit. Generalist surfaced adjacent soundness; a 50-line halo2 wrapper got GPT-5 to the exact fix mechanism — blind. Honest receipt of where the gate works and where it doesn't.

  read case study →
• highsecurityMarch 2026

  OpenClaw CVE-2026-31998 — synology-chat auth bypass

  HIGH CVE on the fastest-growing OSS project in GitHub history — GPT-5 named the exact vulnerability; unanimous gate fired

  read case study →
• highsecurityFebruary 2026

  Moonwell MIP-X43 oracle bug

  1ドル.78M incident — AntFleet caught a sibling of the exploited cbETH config in the same PR

  read case study →

More case studies will be added as findings mature into full post-mortems.