Skip to content

Latest commit

 

History

History
65 lines (53 loc) · 3 KB

incident-response.md

File metadata and controls

65 lines (53 loc) · 3 KB

Security incident response

Reporting and intake

Use GitHub's private vulnerability reporting form for this repository: https://github.com/Capsize-Games/headlesscode/security/advisories/new. The repository is public and private vulnerability reporting is enabled. If that form is unavailable, open an issue tagged security and include only a brief description asking for a private follow-up route. Do not put exploit details, credentials, user data, or other sensitive evidence in a public issue.

Record the report date, affected component and versions or commit, conditions needed to reproduce the behavior, observed impact, and any known workaround. Ask the reporter what contact method and disclosure timing they prefer. Do not promise a response time or a particular outcome.

Triage and containment

Reproduce the report in an isolated workspace with test data. Determine which versions, configurations, and execution paths are affected, and whether the behavior could expose data, run unintended commands, or change persistent state. Keep the report's evidence and reproduction material access-limited; remove credentials and personal data from copies used for testing.

When an active risk is confirmed, limit exposure while a fix is prepared. This can include disabling the affected feature or workflow, pausing affected jobs, restricting access to affected artifacts, and publishing a temporary workaround. Preserve the relevant logs, configuration, and artifacts in a restricted location before cleanup or rebuild. Do not copy secrets into an issue, pull request, test fixture, or public advisory.

Fix, notification, and disclosure

Prepare and review a fix, then verify it against the reported behavior and nearby regression cases. Identify affected versions and the fixed version from the repository tags and package metadata. Notify affected users through the project's release and security advisory channels with the affected versions, impact, available mitigation, and upgrade path. Coordinate advisory timing and technical detail with the reporter; publish enough information for users to assess and address the risk without exposing user secrets or unrelated private evidence.

Post-incident review

Record the timeline, affected versions and configurations, impact, containment, fix, user notification, and disclosure decision. Review which controls failed or were missing, and whether runtime behavior or configuration should change. Record implementation details that changed agent behavior, including prompt and tool changes, model routes, runtime flags, policy and image digests, and control configuration. Assign concrete follow-up work and verify those changes before closing it.

Report template

Summary:
Affected component and versions/commit:
Required configuration or conditions:
Reproduction steps (use test data):
Observed behavior and potential impact:
Workaround, if known:
Evidence available (do not attach credentials or user data):
Preferred contact method:
Disclosure timing or coordination preference: