Yesterday’s deployment failed. The error is in a log file, the rollback plan is in a separate document, and release approval came through in chat. This morning, you need to know what was tried and where to pick up.

I built OKF Todo, a free, open-source Windows task manager, to keep that kind of work together. The example below is a sample task called Fix failed production deployment. Four of seven steps are complete. The deployment log and rollback plan are attached, and a comment in the Timeline records release approval.

OKF Todo showing a failed deployment task with four of seven checklist steps complete, supporting files, and a release approval comment in the Timeline

Give the investigation a place to live

Select New task and give it a name you’ll recognize tomorrow. In the task body, write down what failed and what you know so far. In this sample, the deployment stopped because it couldn’t resolve a secret reference.

Under Attachments, select Add file and attach the deployment log and rollback plan. You can open both from the task the next time you need them.

Keep track of what’s left

Use the checklist for steps you can mark complete. In the screenshot, the deployment output has been captured, the production and staging variables have been compared, the missing reference has been identified, and the rollback package has been validated.

Three steps remain:

  1. Deploy the corrected configuration.
  2. Run production smoke tests.
  3. Update the incident record with the outcome.

When the release manager approves the deployment, use Add comment in the Timeline to record the decision. The sample comment reads, “Received release manager approval; work can continue.”

Pick it back up the next morning

Open the task and look for the first unchecked step. Here, it’s Deploy the corrected configuration. If you need to know why you can proceed, the approval is in the Timeline. If you need to check the original error, the log is still attached.

You can try this without making up an incident. Install OKF Todo, choose Explore with sample data on first launch, and open Fix failed production deployment. Windows install options

Optional: draft the checklist with your own AI assistant

If you already use an MCP-compatible assistant, you can give it sanitized incident notes and ask it to draft a checklist. For example:

Read these incident notes and propose a checklist for investigating the failed deployment. Show me the proposed steps and flag anything you’re unsure about. Wait for my approval before saving anything.

Review the draft, approve the steps you want saved, and ask the assistant to read them back. Then open the task in OKF Todo and check what was saved. MCP setup guide

When you pick up an investigation the next morning, what’s the first thing you usually have to reconstruct?

AI disclosure: Codex helped draft this walkthrough from the project documentation and sample task.