Skip to content

Security: rsitech-ai/crypto-intelligence

Security

SECURITY.md

Security Policy

Supported versions and channels

The repository is pre-release. No stable release, beta, canary, or development snapshot is currently promised security support. Private reports are still accepted and triaged. When a supported release channel is opened, this table must be updated before distribution; support is not inferred from a tag or build artifact.

Channel Security fixes
Stable None released
Beta or canary None released
main and development snapshots Best-effort investigation only

Private reporting

Use GitHub's private Report a vulnerability form. Do not open a public issue for a suspected vulnerability. Do not disclose credentials, exploit details, private keys, account identifiers, or proprietary data publicly. Include the affected revision or version, impact, reproduction conditions, and a private contact route where possible.

Severity and response

Reports are triaged as Critical, High, Medium, or Low using exploitability, required access, confidentiality/integrity/availability impact, local financial authority, and recovery cost. CVSS may inform but does not replace product impact. Maintainers work on a best-effort basis; there is no guaranteed response or remediation SLA. Receipt, severity, fix timing, and disclosure timing are confirmed privately when maintainer capacity permits.

Coordinated disclosure and advisories

Keep details private until a fix or documented mitigation is available and the coordinated disclosure date is agreed. A published security advisory names the affected versions, first patched version, impact, mitigations, and credit when requested. Advisories are published only after the referenced fix and release artifacts are available. GitHub's private vulnerability reporting and repository security-advisory workflow are the canonical coordination channels.

Product authority boundary

Version 1 is local-first research and monitoring software. It has no trading, withdrawal, account-mutation, hosted-inference, or remote-telemetry authority. Public/read-only market data and configured loopback nodes are the only potential external boundaries, and each must be explicitly owned and bounded. See ADR 0003.

Security-sensitive assets include event integrity, exact books, point-in-time features and labels, calibrated probabilities, model packages, ledgers, local session secrets, backups, and update artifacts. Critical controls include bounded parsers and queues, WAL-first durability, authenticated loopback RPC, purpose-scoped egress, atomic writes, redaction, and deterministic evidence.

Release artifact integrity

There are no supported release artifacts yet. A future supported release must publish SHA-256 checksums and signed artifacts bound to the source revision, plus installation and rollback evidence. The advisory or release notes must identify the signing method and patched version; a signature claim is invalid without a verifiable signer identity and artifact digest.

There aren't any published security advisories