TWBG_Materials/文档润色流和知识库构建流/claude-scholar-upstream/skills/nature-response/tests/evaluation-summary.md

Evaluation summary

nature-response is evaluated with synthetic Markdown fixtures. These tests are not executable unit tests; they are behavior contracts for manual and agent review.

Status rationale

Recommended status: Beta.

Rationale:

• The core rules are defined in SKILL.md and modular references.
• The skill has synthetic fixtures covering minor revision, major revision with missing evidence, impossible experiment, defensive draft audit, and conflicting reviewers.
• Each fixture includes expected behavior, forbidden behavior, and pass/fail criteria.
• The examples show expected output shape without using real confidential reviewer comments.
• The skill has not yet been validated on real anonymized revision packages, so Stable would be premature.

Fixture coverage

Fixture	Coverage	Key failure prevented
minor-revision.md	stable IDs, minor comments, missing citation metadata	fabricated citation or line numbers
major-revision-missing-evidence.md	validation request, statistical details, missing evidence	invented results or p-values
impossible-experiment.md	out-of-scope longitudinal evidence	time/funding excuse or fabricated survival data
defensive-draft-audit.md	hostile draft language, vague compliance	accusatory reviewer language
conflicting-reviewers.md	editor priority and incompatible reviewer requests	contradictory manuscript promises

Manual evaluation checklist

• Every fixture has input, expected behavior, forbidden behavior, and pass/fail checklist.
• No fixture uses real reviewer comments.
• Examples are synthetic and do not contain confidential review content.
• Status remains below Stable until real anonymized cases are reviewed.

Promotion path to Stable

Promote from Beta to Stable only after:

• at least two real anonymized revision packages are tested with author permission;
• no fabricated actions, line numbers, statistics, or citations are observed;
• Chinese-note workflows produce usable English response drafts and Chinese confirmation notes;
• edge cases such as conflicting reviewers and impossible experiments remain traceable.