The AI Work Assessment v8 · free · open source · runs in your own coding agent

See the work you've actually done.

Run one prompt in your own coding agent. It reads your local session history, your GitHub activity, and optionally your LinkedIn career history, then writes one HTML file to your machine and tells you where it saved it: what you build, how you verify it, where you fit on a project, and what the record does not establish.

Run the assessment → See a sample profile Star on GitHub
Client details stay outClient names and identifying details are abstracted, and Overflow rejects common credential patterns on upload. Open sourceMIT-licensed. Read the prompt or the code on GitHub before you run it.

run the assessment

Pick the evidence it can use.

// AI Work Assessment prompt Read base prompt

Where do you use AI coding tools?

Choose the footprint you want this assessment to cover. Overflow never connects to or controls another machine.

Choose what the assessment can use.

Leave every source on for the fullest view. Each one carries its own dated coverage window, and activity volume is never treated as a skill score.

5 of 5 selected
Local by defaultSession history is analyzed on your computer. Scrubbed before displaySecrets, personal material, client names, and private project names are removed or abstracted. You decide to shareThe prompt writes one HTML file and stops. Nothing submits itself.
The prompt below updates as you change sources.
Loading prompt…

Paste it into Claude Code, Codex, or any capable coding agent. It finishes by writing a single HTML file to your machine and tells you the exact path.

what you get

Concrete reasons for a project lead to staff you, and limits are mandatory.

agent footprint

Observed session volume, comparable Claude and Codex use, work cadence, and placement in the submitted Overflow cohort.

expertise

Industry and subject-matter knowledge inferred from both career history and the work found in agent sessions.

project decision

The strongest fit, where to add a specialist, and what the available evidence does not show.

See the current sample profile → How matching works →

open source

Don't trust us. Read it.

The current release is Assessment v8. It writes profile schema v9, and its optional multi-environment workflow uses open-source helper v8.0.0. A fixed renderer turns the validated profile data into the final page. The prompt, validator, and renderer are MIT-licensed on GitHub. Run it, fork it, or self-host it without ever talking to Overflow. The version history explains every current number and which older reports remain compatible.

optional: submit it to overflow

The report is yours either way. Sending it does two things.

Your report gains a comparison against the aggregate, unnamed cohort, computed only after submission from the snapshot you sent. It also puts you in the standing community, private by default, where client work gets routed to the people whose evidence fits it.