One task, from the request to the verdict. The agent is never asked to cooperate: every step below is driven by a hook event it was going to fire anyway, and the evidence is read out of command output it was going to produce anyway. Read top to bottom — session and claim, then the edits that grow the task's own scope, then the command results that become evidence bound to exact bytes, then the stop where a state is computed rather than believed.
The system in tiers. core/ never imports the plugin, and the plugin is deliberately thin — everything interesting is importable, and therefore testable. hooks.json is generated from wiring.py, because the scope guard once shipped completely inert while every unit test passed. eval/ is a separate authority: it grades in an evaluator-owned workspace from an exported patch, never inside the candidate's, and it shares none of the runtime's judgement.
Read each lane left-to-right. These lanes describe the shipped runtime and evaluation mechanisms, with their current limits. Everything in the Overview tab maps to real code today. The Planned tab separately explains unfinished milestones, future research, conditional work and proposals that contributors can help develop.
Open a card for remaining work, acceptance criteria and a small first contribution. Sourced from our plans, with unfinished delivery, research and proposals clearly distinguished from shipped behavior.
No matching work. Clear the search or choose a broader topic or status.