DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to plan state handoff between agent execution runs with Harness Engineering

Run the Chapter 2 state notebook to see how agent state carries across execution boundaries.

Short answer: Open the README and find the Chapter 2 table under Running the Notebooks; then Click the Open In Colab badge for State Across Execution Boundaries; run the notebook cells in order and note what state is saved at each boundary. All steps.

Who it is forA developer designing a multi-step agent whose runs are split across sessions or processes and lose context between them.

The problemWhen an agent stops and restarts, important state can be lost, and the failure only shows up later in the workflow.

The outcomeThe developer documents the state the agent must keep between runs and the boundaries where that state is saved and restored.

What you need first

  • A web browser and a Google account to open notebooks in Google Colab
  • A GPU runtime in Colab, since the README notes some notebooks may need one
  • Basic familiarity with Python and Jupyter notebooks

Steps

  1. Open the README and find the Chapter 2 table under Running the Notebooks.
  2. Click the Open In Colab badge for State Across Execution Boundaries.
  3. Run the notebook cells in order and note what state is saved at each boundary.
  4. Record which state is needed to resume the agent and where it lives in the notebook code.
  5. Write a short state plan for your own agent that lists each boundary and its required state.

Prompt for your agent

Using the Nicolepcx/harness_engineering repo, open ch02_state_across_execution_boundaries.ipynb in Colab, run it, and list the state saved at each execution boundary. Apply that list to <placeholder: description of your multi-step agent> and return a state plan with one row per boundary. Verify that each row cites a cell in the notebook.

Caveat. The README gives no detail on what state the notebook saves, so the state plan has to be built from the notebook's own code and output.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for Nicolepcx/harness_engineering and "Plan state handoff between agent execution runs"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer