Changelog
| Version | Date | Changes |
|---|---|---|
| 5.1.0 | 2026-05-20 | Synthesis agent gains an explicit Three-Step Synthesis Protocol with cross_reviewer_quantifier (any/majority/all) and a Forbidden Operations list; the Review Standard must-read list now loads editorial_decision_standards.md and quality_rubrics.md at deep-review time; TROUBLESHOOTING.md adds F1-F8 review-quality failure paths; new revision_coach_agent parses reviewer letters in any format; MODE_GUIDE adds an Auto-Detection block for re-audit triggers; SUBAGENT_TEMPLATES gains lane-specific focus blocks |
| 5.0.0 | 2026-05-20 | Aligned skill version with project-wide pyproject.toml release. Unified frontmatter schema (metadata.* subtree, last_updated, argument-hint) across all five academic writing skills; tightened skill description and trigger-word coverage |
| 4.5 | 2026-04-27 | Added script-backed PRESUBMISSION mechanical audit layer; integrated it into quick-audit, gate, re-audit, and deep-review Phase 0; added pre_submission_readiness lane for full/editor deep-review; documented severity mapping and PDF/source differences |
| 3.0 | 2026-03-16 | Literature search engine (Tavily + S2 + arXiv); 9-dimension ScholarEval with Literature Grounding (12%); weighted-plus scoring model (weighted average + interaction/penalty terms); Literature Reviewer agent; PDF metadata extraction; 3 new eval prompts |
| 2.0 | 2026-03-11 | Full rewrite: venue filtering, multi-perspective review agents, re-audit mode, templates, examples, quality rubrics |
| 1.0 | 2026-03 | Initial version: 4 modes, script-based audit, 4-dim + 8-dim scoring |