This repository is a public, tamper-evident commitment ledger and its independent verifier. We take reports about its integrity seriously.

• A flaw in verify.py or verify_bitcoin.py — e.g. a way to make verification pass for predictions that were never committed, or fail for ones that were.
• A weakness in the anchoring / manifest protocol (see METHODOLOGY.md) — for example, a way to forge or alter a manifest_hash, content_hash, or verifier_sha256 binding.
• A way to bypass the append-only integrity guard (.github/workflows/integrity.yml).
• A suspected integrity problem with a published anchor, model registration, or .ots proof.

Email security@offensiveedge.com with details and, where possible, a reproduction. Please do not open a public issue for a suspected vulnerability until we have had a chance to respond.

We aim to acknowledge within 3 business days. Confirmed integrity incidents are then disclosed publicly under the policy in incidents/README.md (7-day window, append-only, no silent edits).

In scope — the public surface of this repository:

• verify.py, verify_bitcoin.py
• the manifest / anchor protocol and methodology
• the integrity CI guard
• the published ledger artifacts (anchors/, models/, reports/, .ots proofs)

Out of scope — EdgeSeeker's private prediction pipeline and infrastructure. The predictions, salts, and model binaries are not in this repository (see OPERATIONS.md); issues there are not part of this public repo's attack surface.

• "The model was wrong about a game." That is performance, not a vulnerability — see the performance reports and incidents/README.md → "What does not qualify as an incident."
• Verification failing because you altered a tracked file locally (including a line-ending rewrite — see .gitattributes). That is the verifier working as intended.