Module grade 

Stage 4 — grade.

Decomposed into focused units:

• transcript_check — grade a transcript_check assertion (regex over the run’s tool invocations).
• judge_tasks — emit LLM judge tasks + the skill-invocation meta-check (emit_judge_tasks), the default mode.
• [finalize] — fold judge responses + transcript checks into grading.json (--finalize mode).

Both modes operate over a shared GradeContext assembled by the CLI.

Re-exports

pub use finalize::FinalizeSummary;

pub use finalize::finalize;

pub use judge_tasks::EmitSummary;

pub use judge_tasks::check_skill_invoked_from_transcript;

pub use judge_tasks::emit_judge_tasks;

pub use transcript_check::grade_transcript_check;

Modules

finalize

Grading finalize.

judge_tasks

Judge-task emission.

transcript_check

Transcript-check grading.

Structs

GradeContext

The resolved inputs both grade modes read: the iteration directory, the conditions manifest, and the validated evals config (its skill_name is the one used in meta-check rubrics).