ITU-T Recommendation Reviewer

A model-agnostic AI agent skill that reviews draft ITU-T Recommendations (including ITU-T | ISO/IEC common texts, amendments, and corrigenda) for compliance with the official ITU-T drafting rules — before the text goes to TSB or is proposed for consent, determination, or approval.

The skill is written entirely in plain Markdown, so any large language model or AI agent can use it — load the files into the model's context, or install them in any agent framework that supports the open Agent Skills format (SKILL.md + reference files).

It encodes the drafting and presentation rules from three authoritative sources:

ITU-T Editing Guidelines — Author's guide for drafting ITU-T Recommendations (06/2023) — the primary rulebook for pure ITU-T texts (ITU permalink)
Rules for presentation of ITU-T | ISO/IEC common text (09/2014) — overrides the Author's guide for joint ITU-T | ISO/IEC documents
Recommendation ITU-T A.1 (09/2019) — ITU-T working methods (ITU permalink)

The ITU source documents themselves are not redistributed in this repository. They are © ITU and available free of charge from the links above and the ITU-T website.

What it does

Given a draft Recommendation (as .docx, PDF, or pasted text), the skill instructs the model to:

Classify the document type first — pure ITU-T text, ITU-T | ISO/IEC common text, or amendment/corrigendum — and apply the correct rulebook.
Run a six-pass review: skeleton & mandatory elements → boilerplate wording → references & citations → definitions & abbreviations → figures/tables/equations/notes → quality checklist & language.
Produce a structured review report with an overall verdict (READY / MINOR REVISION / MAJOR REVISION), findings graded Critical / Major / Minor, each pinpointed to a clause and citing the governing rule, plus the Annex D pre-approval checklist as a pass/fail table.
Offer follow-through — corrected clauses with revision marks, rewritten definitions, or ready-to-paste boilerplate.

The skill checks drafting/presentation compliance and internal consistency. It does not judge the technical correctness of the standard's subject matter, and it never predicts approval outcomes.

Installation

Any LLM (ChatGPT, Gemini, Qwen, DeepSeek, local models, …)

The skill is just Markdown. Attach or paste SKILL.md plus the five files in references/ into the conversation (or your tool's knowledge/context feature), then ask the model to review your draft following SKILL.md.

Agent frameworks supporting the Agent Skills format

Clone this repository into your agent's skills directory, e.g. for Claude Code:

git clone https://github.com/TheTaoism/itu-t-recommendation-reviewer.git \
  ~/.claude/skills/itu-t-recommendation-reviewer

Claude apps (claude.ai / desktop / Cowork)

Download itu-t-recommendation-reviewer.skill and add it in Settings → Capabilities → Skills.

Usage

Upload or paste your draft and ask, for example:

"Review this draft Recommendation for ITU-T compliance."
"Does my draft follow the ITU-T Author's Guide? Prepare it for TSB submission."
"检查这份 ITU-T 建议书草案是否符合编辑规则。"
"How should I structure the Definitions clause of an ITU-T | ISO/IEC common text?"

The report is delivered in your working language; quoted rule text and suggested replacement text stay in English (the publication language).

Repository layout

SKILL.md                          # Skill entry point: workflow, report template, severity calibration
references/
  structure-rules.md              # Mandatory elements, clause order, required boilerplate
  formatting-rules.md             # Fonts, numbering, figures, tables, notes, citation style
  definitions-rules.md            # Rules for writing definitions
  common-text-rules.md            # Deviations for ITU-T | ISO/IEC joint texts
  quality-checklist.md            # Annex D pre-approval checklist + A.1 process checks
itu-t-recommendation-reviewer.skill   # Packaged skill for Claude apps (optional convenience)

Contributing

Contributions are welcome — see CONTRIBUTING.md.

Please note: all changes go through pull requests and require review and approval by the maintainer before they are merged. Direct pushes to main are not accepted.

Disclaimer

This is an independent community project. It is not endorsed by, affiliated with, or approved by the International Telecommunication Union (ITU), nor by any AI vendor. "ITU" and "ITU-T" are used solely to describe the subject matter. Always verify review findings against the current official ITU-T editing guidelines.

License

MIT — anyone may download, use, modify, and redistribute this skill, commercially or otherwise, provided the license notice is preserved.

Itu T Recommendation Reviewer

README