The problem
Self-reviews and promotion cases are written from memory, months after the work. What gets written is what the author recalls and can phrase well, not what the record shows.
What it does
Impact Advocate interviews you about your role, inspects the GitHub history you have authorized, and writes a Markdown evidence document. It separates contributions it observed from inferences it made, and where GitHub holds no evidence it says “Not visible in GitHub” instead of filling the gap. Every claim is linked or counted.
It runs inside your own Claude session with your own access. There is no server, no telemetry and no extra AI provider. It is self-directed and does not evaluate colleagues.
Who it’s for
Data analysts, analytics engineers, data scientists, senior and staff engineers, engineering and analytics managers, open-source maintainers, and Claude Code users.
Join the beta
Run it on your own history for a timeframe you choose, against your job description, OKRs or career ladder. Check the claim links, confirm what matters, and challenge anything unsupported. Then submit five ratings and three evidence examples (one found correctly, one missed, one questionable) through the feedback template.
Feedback issues are public, so don’t paste generated packets, private repository data, credentials, or details that identify colleagues.
Beta page · How-to and install · Submit feedback · System specification