Tricorder / open technical beta

Can an AI-generated promotion packet survive a technical practitioner's scrutiny?

Impact Advocate turns your own GitHub history into a self-review draft, linked to evidence and calibrated to your role. We need skeptical engineers and analysts to discover what it gets wrong.

Try Impact Advocate Submit your feedback
Who we're looking for

Data analysts, analytics engineers, data scientists, Senior/Staff software engineers, engineering and analytics managers, open-source maintainers, and Claude Code users. Solo and multi-contributor repositories are both useful.

What it does: interviews you about your role, inspects your authorized GitHub history, and produces a Markdown evidence document. It distinguishes observed contributions from inferences and says "Not visible in GitHub" when evidence is unavailable.

How to test it

  1. Read the how-to and installation instructions.
  2. Run it on your own GitHub history, using a timeframe and role expectations, job description, OKRs, or career ladder you select.
  3. Inspect the result: check claim links, confirm important work, and challenge unsupported interpretations.
  4. Submit the five ratings and three evidence examples (correctly found, missed, questionable).
Five things we'd like to learn
  1. Could you install and complete it?
  2. Did it find the most important work?
  3. Were its claims accurate?
  4. Did its ratings fairly reflect your role expectations or ladder?
  5. Would you use the final document?

Ratings are from 1 to 5, with N/A available.

Prefer to talk it through?

Want to discuss your experience, the approach, or an edge case? Book an optional 30-minute coffee conversation.

Book a 30-minute coffee

Privacy and scope

Privacy-first: runs entirely inside your own Claude session, with your own access. No server, no telemetry, no extra AI provider. On an employer repo, check that your company's policy allows your Claude plan to read its code.

This beta is self-directed and does not evaluate colleagues. Feedback issues are public. Do not paste generated promotion packets, private repository data, confidential details, credentials, or colleague-identifying text into GitHub Issues. Describe examples at a safe level of abstraction.

See the system specification for the evidence model and known limitations. Report installation bugs via GitHub Issues.