Tooltip text
Editing Agent Scoring Sub-Agent 1 LOOP INITIAL SKILL.md Generate Prompts Baseline Score Target Lowest Apply One Fix Re-score Keep or Revert results.tsv Final Version

Controls

State Inspector

System Idle
Click any node in the diagram to inspect its function in the optimization pipeline.

Evaluation Rubric

Clarity 92
How easily the AI's output can be understood by the user.
Accuracy 88
Factual correctness and adherence to ground truth.
Completeness 85
Whether all parts of the user request were addressed.
Conciseness 81
Avoiding unnecessary verbosity or filler text.
Tone 64
Adherence to the requested persona or brand voice.
Formatting 90
Proper use of markdown, lists, and structural elements.
Examples 78
Quality and relevance of provided examples.
Edge Cases 72
Handling of ambiguous or out-of-bounds inputs safely.

Simulate Evaluation

Score < Baseline (80). Action: REVERT