Agents say done. Canon checks the work.
You approve the plan before any code. Then a second agent with no memory of the build grades every acceptance criterion against the code, with the file and line as evidence, and a failing grade blocks the close. Plans, decisions and criteria stay in your repo as plain markdown. Run canon to open the Cockpit that runs it all.
evaluator-run-idProof a separate evaluator run wrote this report.sprint completechecks the run log for it.graded-headThe commit that was graded. If the code moves on, the close blocks until a fresh evaluator grades it.ModelThe grader's model is recorded in the report, and a report without the line can't close.- EvidenceA
file:lineand a run behind each pass. Here the evaluator also broke the code on purpose to check its test would notice.
Things a test suite can't reach.
A test that could not fail.
Across two sprints on a real agentic app, the gates never found a wrong number. They found a check that agreed with itself, safety branches nobody had ever executed, and screenshots timestamped 13 minutes before the commit that replaced the feature.
Each criterion gets a pass or fail with the file and line that back it, and nothing closes on a fail.
Open canon. Two commands. Everything lands in the repo.
Commands
Five things to remember.
A second agent grades the work.
At sprint complete, a fresh agent with no memory of the build checks each acceptance criterion against the code and gives evidence. A failing verdict blocks the close.
Lessons that earn their place.
A sprint leaves a candidate lesson in a queue, not in the agent's memory. A later, separate session judges it and promotes only what will last.
One board per project.
Tickets sit in open, in progress, done and discarded. Start or resume one from its card, in its own worktree, and search across all of them.
See every agent at a glance.
Each session is working, done, idle or needs you, in the sidebar, on cards and on tabs. One that is waiting on an answer says so, so you never hunt for it.
A daemon owns your terminals.
Agents run under a local daemon, not inside the browser tab. Refresh or close the tab and they keep working. Restart the daemon and each interrupted session can be resumed.
Plan. Build. Prove it.
Install once, add the sprint skill to a project, then ask your agent to run sprint start.