AI Research Writing Pipeline

中文说明

Evidence-gated Codex skill for full-cycle academic manuscript writing.

This repository packages a reusable Codex skill that turns a research project, experiment folder, draft paper, or messy multi-agent workspace into a traceable academic-writing pipeline. It is intentionally domain-neutral: the active project supplies the field, venue, method facts, evidence rules, and style constraints.

What It Does

• Maps a research workspace before editing prose.
• Integrates idea evaluation, literature verification, paper reading, baseline selection, evidence audit, writing, polishing, figures, review, rebuttal, data/release statements, and side-output workflows.
• Separates literature, experiment evidence, writing, figures, review, mentor advice, integration, administration, and release work.
• Requires claims, numbers, references, figures, and unknowns to be registered before they are promoted into a manuscript.
• Preserves evidence boundaries, negative results, post-hoc explanations, and unsupported gaps.
• Helps organize chaotic projects without breaking traceability paths.
• Includes a small inventory script for first-pass project cleanup planning.

Repository Layout

ai-research-writing-pipeline/
|-- README.md
|-- LICENSE
`-- skills/
    `-- ai-research-writing-pipeline/
        |-- SKILL.md
        |-- agents/
        |   `-- openai.yaml
        |-- references/
        |   |-- artifact-contracts.md
        |   |-- capability-playbooks.md
        |   `-- framework-patterns.md
        `-- scripts/
            `-- workspace_inventory.py

Install

Install the skill into Codex from this GitHub repository:

python3 ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py \
  --repo youngandboor/ai-research-writing-pipeline \
  --path skills/ai-research-writing-pipeline

Restart Codex after installation so the new skill is discovered.

Use

Invoke the skill explicitly:

Use $ai-research-writing-pipeline to inspect this research project, build claim
and evidence ledgers, and turn the current draft into a traceable manuscript
workflow.

For a messy existing project, start with the bundled inventory script:

python3 ~/.codex/skills/ai-research-writing-pipeline/scripts/workspace_inventory.py .

Design Principles

This skill treats academic writing as controlled integration, not as free-form prose generation.

The main discipline is:

1. Orient the project.
2. Classify the request into a lane.
3. Freeze method and evidence invariants.
4. Build evidence ledgers.
5. Register claims and numbers.
6. Draft only from registered evidence.
7. Render figures from source data.
8. Integrate once through a controller role.
9. Red-team from the current artifact.
10. Repair by expected score gain and evidence risk.
11. Gate final readiness through validators.

Included References

• framework-patterns.md distills reusable patterns from state-machine research workflows and multi-agent paper-production systems.
• artifact-contracts.md gives lightweight JSON/JSONL contracts for task packets, claim registries, numeric claims, reference ledgers, figure manifests, and completion manifests.
• capability-playbooks.md integrates the common capabilities of idea evaluators, literature tools, paper readers, baseline selectors, writing and polishing workflows, figure agents, reviewers, rebuttal planners, data statement helpers, presentation builders, and release gates into this single skill.

License

MIT License.