open-bridge orchestrates AI agents and handles configuration files plus references to secret stores (KeyVault, 1Password, Keychain URIs — never the secrets themselves). A flaw in how those references are parsed, in agent-dispatch logic, or in a shipped skill or script can have real consequences. We take such reports seriously and appreciate the time it takes to file a good one.

Please do not open a public GitHub issue for security bugs. A public report gives an exposure window before a fix exists.

Instead, email info@bks-lab.com with SECURITY in the subject line. Include as much of the following as you can:

• Affected component — the file, skill (skills/<name>/), script (scripts/), protocol, or agent-dispatch path involved.
• Reproduction — minimal steps, config, or input that triggers the issue. A redacted bridge-config.yaml (secrets removed) helps.
• Impact — what an attacker can read, write, exfiltrate, or execute, and under what preconditions.
• Version — the commit SHA or release you tested against.

If you have a suggested fix, include it — but a clear report alone is plenty.

In scope — the Bridge's own code and logic:

• Skills, scripts, and protocols shipped in this repository.
• Agent-dispatch logic — how agents are spawned, scoped, and coordinated.
• Config handling — parsing of bridge-config.yaml, cluster-wrapper YAML, and the resolution of secret-store references (KeyVault / 1Password / Keychain URIs).
• Template, schema, and validation logic that other instances depend on.

Out of scope:

• A user's own private user/ branch content — personal configs, notes, and instance-specific data live outside this repository's responsibility.
• Third-party agent runtimes — Claude Code, Gemini CLI, and other host harnesses. Report those to their respective vendors.
• Secrets the user committed themselves. open-bridge stores references, not values; a raw credential placed in a tracked file by the operator is an operational mistake, not a vulnerability in this project. (We still recommend rotating it immediately.)

If you are unsure whether something is in scope, report it and let us decide.

open-bridge is a community open-source project maintained on a best-effort basis. We want to set honest expectations:

• We aim to acknowledge a report within a few business days.
• There is no SLA and no guaranteed fix timeline.
• There is no bug bounty — we cannot offer payment for reports.

We will keep you informed as we triage and, where appropriate, credit you in the fix (only with your consent).

We ask that you give us a reasonable window to investigate and ship a fix before disclosing the issue publicly. We will work with you on timing and will not take action against anyone who reports in good faith, acts within this scope, and avoids privacy violations or service disruption while testing.

Only the latest release and the current main tip receive security fixes. Older commits and forks are not maintained — update to the current main before reporting, and confirm the issue still reproduces there.

Last updated: 2026-06-21.