brainkit's core promise is structural privacy: a person's compiled vault must never contain a note they aren't allowed to read. Reports that break that promise are treated as the highest severity.

Please report vulnerabilities privately — do not open a public issue.

• Preferred: GitHub's private vulnerability reporting — use Security → Report a vulnerability on this repository.
• Alternative: email joe@ticc.net with a description and reproduction steps.

You'll get an acknowledgement within a few days. Please allow time for a fix before disclosing publicly.

In rough priority order:

1. Cross-vault leakage — any way a compiled vault, search index, dashboard, or MCP response can expose content from a space the person cannot read (including via symlinks, path traversal, link stubs, or crash states).
2. Write-scope bypass — a write-back or ingest path accepting a change outside the sender's allowed spaces, or a promotion publishing without human approval.
3. Intake spoofing — getting a note filed as someone else, or frontmatter/ path injection through a capture channel — including the webhook receiver: signature forgery or verification bypass, replay past the dedup window, bypassing a source's rate limit, or routing a delivery to a person the source shouldn't reach.
4. Dashboard issues — CSRF, XSS, or any action reachable without the guarded primitives.

Out of scope: vulnerabilities in your embedding provider, agent runtime (Claude Code, Hermes), or deployment infrastructure — though we're happy to document mitigations.

brainkit is pre-1.0; only the latest release (and main) receives security fixes.