See who's steering the AI.
Sus Coder shows how your engineers work with Claude Code: who leads with their own judgement, and who hands everything over and accepts what comes back.
Know how your team really works with AI.
Not how much they use it. Whether they still bring their own understanding to the work.
One command to install
A small hook joins Claude Code on Mac, Linux and Windows. Nobody changes how they work.
Secrets stay on the laptop
Keys, passwords, emails and home folders are removed before anything leaves, then checked again on the server.
Six dimensions per prompt
Specificity, clarity, verification, own diagnosis, catching mistakes and adding their own thinking, read the same way every time.
A map, not a leaderboard
Each engineer lands as a Pilot, Checker, Delegator or Passenger, with a range and a confidence level instead of a single score.
Leaked keys, caught
Credentials typed into prompts are flagged, even with typos, while placeholders and documentation examples are left alone.
Ask in plain words
Type or talk: "How was Rupesh's last session?" The assistant answers from the same data as the dashboard.
From a prompt to a profile in three steps.
Capture
The hook sends each prompt, tool call and reply ending, already redacted, to your own server.
Read
Every prompt is scored against a fixed codebook, and each turn's tokens and cost are recorded.
Place
Sessions add up to a profile: where each engineer sits on the map, how sure we are, and the prompts behind it.