Incident response runbook

Use this when the project may have shipped a bad release, leaked a secret, or published misleading security or compatibility information.

Stabilize

1. Stop the action that is making the incident worse.
2. Assign one maintainer as incident lead.
3. Create a private timeline with times, actions, and links.
4. Preserve logs, workflow runs, artifacts, and reports.

Contain

• Bad package release: yank or deprecate the release if the registry supports it.
• Leaked secret: revoke or rotate the credential immediately.
• Broken docs: disable deployment or publish a correction.
• Broken CI automation: disable only the failing automation that creates risk.

Fix

1. Create the smallest safe fix.
2. Add a regression test or validation check.
3. Run make check and any focused reproduction.
4. Prepare release notes or an advisory when users need action.

Recover

1. Publish the fix.
2. Verify the fix from a clean environment.
3. Communicate impact and mitigation.
4. Update the relevant runbook with the missed step.