Digital Employee Works

English

简体中文

English

简体中文

Appearance

Research Report Production Engine V1.3 Quick Start ​

1. Download ​

Download ZIP ​

Download the current source ZIP

Extract it and enter the joinwell52 directory.

Use Git ​

bash
git clone https://github.com/joinwell52-AI/joinwell52.git
cd joinwell52

2. What the download contains ​

text
research/skills/
  Research Skills V2.0 and the three intelligence profiles

research/intelligence/
  AI Platform, GitHub Engineering, and Published Research source registry

research/runtime/
  Scheduler, Worker Contract, Runtime Record, column-plan, and candidate specifications

scripts/
  Runtime, Research Intelligence, and website data validators

.github/workflows/
  GitHub scheduling, validation, and Pages deployment workflows

docs/
  Bilingual VitePress website

3. Prerequisites ​

The website and validators require:

  • Git;
  • Node.js 22;
  • npm;
  • a writable GitHub repository;
  • a ChatGPT environment for the actual research work.

4. Install dependencies and validate ​

bash
npm install
npm run intelligence:validate
npm run runtime:validate
npm run docs:build

When these commands pass, the source registry, Runtime records, three-column plan, Publication Candidate batch, and website satisfy the current repository contracts.

5. Preview the website locally ​

bash
npm run docs:dev

Or:

bash
npm run docs:build
npm run docs:preview

6. Configure your three research columns ​

The default definitions live in:

text
research/runtime/SCHEDULER.json
research/intelligence/REGISTRY.json

The default columns are:

  • Digital Employee;
  • Industry Architecture;
  • Open-source Engineering.

You may replace the labels, research questions, and sources, but preserve these rules:

  • every selected object has exactly one primary column;
  • every column receives Selected or No Selection each day;
  • non-selection records coverage and a concrete reason.

7. Configure source intelligence ​

Edit:

text
research/intelligence/REGISTRY.json

It contains:

  • OpenAI, Claude, Gemini, Cursor, GitHub Copilot, and Microsoft Copilot Platform;
  • controlled GitHub organizations, repositories, and query topics;
  • arXiv, OpenReview, AI research labs, conferences, and publication sources;
  • evidence levels, scan frequency, and cross-pipeline deduplication rules.

8. Create the workers in ChatGPT ​

The GitHub workflow opens execution slots; it does not independently complete research writing.

Use this file to create the corresponding ChatGPT workers:

text
research/runtime/WORKER-PROMPTS-V2.md

Formal shifts:

text
09:00 Engine
10:00 Queue + Research Intelligence
11:00 Knowledge
Monday 12:00 Architecture
15:00 Production
20:00 Publication
Sunday 20:30 Weekly
Wednesday 10:00 Academic

Every worker must:

  1. read the authoritative upstream artifacts;
  2. execute only its assigned responsibility;
  3. write a structured work-outcome report;
  4. update the Runtime Log;
  5. commit durable outputs;
  6. fetch the Git commit and verify the result.

9. Run 10:00 intelligence and three-column triage ​

The Queue worker executes:

text
Skill 01-P AI Platform Change Intelligence
Skill 01-G GitHub Engineering Intelligence
Skill 01-R Published Research Intelligence
↓
Cross-source deduplication
↓
Skill 02 Three-Column Research Triage

Outputs:

text
research/intelligence/runs/YYYY/MM/YYYY-MM-DD-intelligence.json
research/runtime/plans/YYYY/MM/YYYY-MM-DD-plan.json

10. Run 15:00 complete-report production ​

The Production worker reads eligible research objects and executes:

text
Skill 05 Research Writing
→ Skill 06 Visualization
→ Skill 07 Evidence & Citation
→ Skill 08 Publication Editing

It creates complete bilingual Publication Candidates at:

text
research/runtime/candidates/YYYY/MM/YYYY-MM-DD-candidates.json

Production does not place the report in the public article directory.

11. Run the 20:00 release shift ​

The Publication worker consumes complete candidates only:

text
Publication Candidate
→ public bilingual Markdown
→ metadata / indexes / website
→ GitHub Commit
→ Commit Verify
→ Release

It must not search for new sources, repeat analysis, or write an article from scratch at 20:00.

12. Inspect operations ​

Operations Center path:

text
/en/runtime/

It reports:

  • coverage of all three intelligence pipelines;
  • topic decisions for all three research columns;
  • actual work outcomes for every scheduled shift;
  • 15:00 Publication Candidates;
  • 20:00 release results;
  • Runtime Records and GitHub verification.

13. Important boundaries ​

  • ChatGPT performs the actual worker execution;
  • a GitHub Actions trigger does not prove that work completed;
  • Queue cannot formally complete without a complete intelligence run;
  • Publication cannot release without a complete Publication Candidate;
  • a result without a Runtime Record and Commit Verify is not an official output;
  • the repository is publicly downloadable, while reproduction, adaptation, redistribution, and commercial use follow the current LICENSE.md.

Return to the product page ​

Research Report Production Engine V1.3

Last updated: