Please do not open a public issue for a security problem.

Report it privately by email to dave@cerebralgardens.com with SECURITY in the subject line. Include enough for us to reproduce it: the plugin and version, the harness (Claude Code or Codex), the backend involved, and the steps or proof-of-concept.

You can expect an acknowledgement within a few days. We will confirm the issue, agree a fix and a disclosure timeline with you, and credit you in the changelog unless you would rather stay anonymous.

Never include a real API token, .flightdirector/secrets.json, or any other live credential in a report. If you believe a token has been exposed, revoke it on your forge first, then tell us.

This project is pre-1.0 and moves fast: fixes land on the latest released version of the affected plugin. There are no long-term support branches.

These plugins run inside an agent session on your own machine and talk to your forge. The things worth reporting:

• Credential handling — anything that could write a token from .flightdirector/secrets.json into a log, a commit, an issue body, a PR, or a request to a host other than your configured backend.
• Command or argument injection — issue titles, branch names, labels, or API responses that escape into a shell command run by the dispatcher or a skill.
• Unintended writes — a path by which a skill pushes, merges, or force-updates a branch without the explicit human gate the workflow requires.

Out of scope: vulnerabilities in Claude Code, Codex, or your forge itself (report those to their maintainers), and anything that requires an attacker to already control your machine or your agent's instructions.