Error
Inputs used
Agent message
Feedback
Add what looks wrong or missing — one point at a time. When you're done, Save stores them with this run, then pick one below to diagnose and fix.
Diagnose & fix
Pick a feedback item; the agent reads the whole run — hard rules, what it placed, the checks, the repairs — then explains what happened and proposes a targeted fix. Apply it to get the next version.
Propose a rule or skill
Send a proposed rule or skill to the Review tab for the team to consider. This does not change this wall or the codebase — it's captured for review only.
Tick the suggestions you want the team to review, then submit.
Runs
Every reinforcement run on this endpoint — shared across everyone. Click one to open it in the Workspace.
Feedbacks
Every run's feedback mapped to its files — the input IFC, each version's IFC, and which feedback produced each version. Download any file; open the run to act on it.
Review
Approved feedback grouped by run — every rule and skill an SME flagged across that run's iterations, in one place. Expand a run to see the full list, or open its artifacts.
Developer
Runs SMEs escalated for a code/rule fix — each carries the feedback, the LLM root-cause analysis, and the input IFC, trace, and reinforced IFC.