doctor makes 2 agent calls and 0 judge calls
ok    config     loaded promptcheckup.toml
ok    prompts    prompts/system.md
warn  cases      20 cases, 0% synthetic; 10 held-out cases, need at least 20; add cases before testing proposals
ok    agent      plain_agent:run ran in 0.001s; output: {"answer": "Hello, Ada!"}
ok    prompt-use the output changed with the supplied prompts
ok    score      plain_agent:score returned 1.0
skip  llm        skipped by --no-llm
ok    budget     gate: 80 calls; check-up: 122 calls
