v2.21.0

AIGC Detector & Rewriter Skill v2.21.0

This release strengthens the skill’s reliability, auditability, and rewrite-control logic for academic and professional text rewriting tasks. The main focus is reducing over-rewriting, improving detector-risk localization, and making the workflow safer when no external detector score is available.

Key Updates

• Added clearer rewrite-unit decision rules to prevent unnecessary full-section or full-chapter rewriting.
• Improved the distinction between sentence-level, paragraph-level, paragraph-group, subsection-level, and chapter-level rewrite scopes.
• Added a No External Detector Score Rule:
  • Internal D1–D17 scores must be treated only as risk-location estimates.
  • The skill must not claim that detector scores have been reduced unless external retesting data is provided.
  • Chapter red-line ledgers should not be created without external detector evidence.
• Strengthened safeguards against unsupported claims, including false “score reduction” claims.
• Improved workflow discipline for cases where users only provide text but no detector report.
• Clarified that chapter-level rewriting means planning and controlled paragraph-group revision, not wholesale regeneration.
• Enhanced consistency between detection principles, rewrite methods, and the main SKILL workflow.

Reliability Improvements

• Reduced the risk of over-editing human-written academic content.
• Improved preservation of the user’s original argument, structure, and evidence.
• Added stricter control over rewrite granularity.
• Improved compatibility with iterative review workflows where the user provides detector feedback later.
• Strengthened the distinction between internal diagnostic estimation and externally verified detector results.

Workflow Changes

• The skill now prioritizes the smallest effective rewrite unit.
• If no external detector score is available, the workflow must:
  • Mark all internal scores as “internal risk estimate only.”
  • Avoid exact chapter-level red-line claims.
  • Recommend external retesting after revision.
  • Rebuild the Chapter Version Ledger only after the user provides external detector results.

Intended Use

This version is designed for controlled rewriting of academic, thesis, research, and professional documents where the goal is to reduce AIGC-style risk while preserving meaning, evidence, authorial intent, and human writing characteristics.

Notes

This release does not guarantee that any third-party detector will produce a specific score. Detector outcomes depend on the external detector, text domain, language, document structure, and retesting conditions.

v2.19.1

AIGC Detector & Rewriter v2.19.1

This release updates the AIGC Detector & Rewriter skill for academic AI-writing-risk diagnosis and minimal-edit thesis rewriting.

Highlights

• Supports English and Chinese thesis text.
• Uses detector-informed paragraph risk diagnosis.
• Processes one chapter per rewrite round.
• Preserves citations, data, hypotheses, statistical results, headings, and document structure.
• Requires external re-test evidence before treating a rewritten version as improved.
• Includes safeguards against full-paragraph regeneration, fabricated content, word-count shrinkage, paragraph merging, and academic-integrity risks.

Recommended Use

Use this skill for academic AI-writing-risk reduction, especially when working with external detector results such as Turnitin AI, CNKI AIGC, GPTZero, or similar tools.

Important Note

This skill does not guarantee a specific detector score. It reduces common AI-writing-risk patterns through structured, minimal-edit revision and requires external re-testing for verification.