Import a produced .py file under a unique module name (no sys.modules reuse).

Resolve a plugin's cache dir portably -- hardcoding one machine's absolute path (e.g. C:\\Users\\<you>\\...) made the ponytail/caveman arms unrepr

scores: {(task_id, label): {SCORE_KEY: int}}. For each task the 'complete' label must strictly out-score the 'stub' label, else the judge (or the

Import a produced module by name with workdir on sys.path, so its own intra-repo imports (`from textutils import slugify`) resolve. Fresh each cal

Each task's good ref must score correct+safe; the bad ref must be caught on its declared axis. Verifies the instruments before any API spend.

Live: the judge model must rank each complete ref above its stub.

Concatenate the agent's source files (tests + artifacts excluded), with name headers.

Commit the seeded repo so we can diff exactly what the agent changes.

Split a produced .py file at the first TOP-LEVEL self-check marker (a `__main__` guard or a demo()/selfcheck() function) through end of file. Retu

Plugin-dir resolution must be portable: env override wins, and a missing install fails loudly (sys.exit) instead of silently passing a non-existen

Return the mode-filtered Ponytail context injected before LLM turns.

LOC of fenced code blocks in a chat answer: (total incl comments, code-only).

LOC over code-extension source files only (generated images/data can't pollute it). total_loc counts every non-blank line including comments and d

Non-blank, non-comment lines of code: fenced blocks, or the whole response when the model emitted bare code with no fence.