Knowledge Systems of Record can contain sensitive institutional information and can influence AI agent behavior. Treat issues involving the following as particularly important: unauthorized corpus access, provenance bypass, malicious source ingestion, prompt injection through knowledge content, privilege escalation, unsafe MCP exposure, build tampering, dependency compromise, and anything that makes ungoverned knowledge appear authoritative.

Report vulnerabilities privately via GitHub Security Advisories. Do not open public issues for security reports.

Please include: the affected version or commit, a reproduction, and the impact as you understand it. You will get an acknowledgement within 72 hours.

• @panaversity/ksor 0.x is pre-release, but it is NOT inert: since 0.0.7 the CLI ingests corpus content, sends it to an embedding provider, writes it to Postgres, and serves it over a network-facing MCP door. In scope accordingly: the MCP surface and its auth postures, the ingest write plane and its row-level-security roles, the audience/takedown filters that decide what a caller may be served, snapshot-token handling, and egress to the configured embedding provider. The attack surface grows with each implemented verb, and this policy applies to all of it.
• Supply-chain reports about this repository's dependencies are in scope — the repo enforces a 48-hour release quarantine (minimumReleaseAge) and a build-script allowlist in pnpm-workspace.yaml.