Skip to content

Quick start

Start with run, which evaluates one commit and makes hook development quick:

bisectrunk run --repo ../subject --at HEAD \
  --run 'test "$(cat "$BISECTRUNK_WORKTREE/marker.txt")" = good'

Once the hook behaves correctly, bisect a known-good/known-bad range:

bisectrunk bisect --repo ../subject --good v1.0.0 --bad HEAD \
  --run 'test "$(cat "$BISECTRUNK_WORKTREE/marker.txt")" = good' \
  --jobs 8

Endpoint verification runs first. Each round evaluates evenly spaced probes in parallel, then narrows toward the leftmost bad result. The final output links to the run directory, logs, JSON report, and Markdown report.

For histories with multiple transitions, scan instead:

bisectrunk scan --repo ../subject --range v1.0.0..HEAD \
  --run './classify.sh' --jobs 8

Press Ctrl-C once to stop dispatching new work. Finished evaluations are flushed; use the printed bisectrunk resume RUN_DIR command to continue.