SETUP_GUIDE.md

Setup Guide

0. Install Required Software

Before cloning the template, read:

• docs/SOFTWARE_AND_PLUGIN_REQUIREMENTS.md
• docs/REAL_PROJECT_OPERATING_GUIDE.md when you want to use the kit on a real dissertation, thesis, manuscript, report, or evidence-synthesis project.

Minimum setup:

• Codex or Codex CLI
• GitHub account
• Git
• Terminal or shell: Apple Terminal, Windows Terminal, Git Bash, or Linux terminal
• Python 3 for the built-in local scripts

Optional but useful:

• GitHub CLI (gh)
• GitHub Desktop
• ChatGPT Codex Connector
• Obsidian
• LibreOffice or Microsoft Word
• ChatGPT or another LLM chat for optional manual external review through scripts/build_external_review_bundle.py
• Claude Code for one optional direct external-review runner through scripts/claude_independent_review.py if you have access

1. Copy Or Clone

Clone the repository or copy this folder to a new project location. Keep it private until the public-release audit is complete.

Recommended project path pattern:

~/Documents/research-agent-starter-kit

2. Fill In Project Basics

Create your own:

• RESEARCH_PROJECT_BRIEF.md
• research-wiki/PROJECT_OVERVIEW.md
• research-wiki/OPEN_QUESTIONS.md
• project-specific requirement notes, such as compliance/PROJECT_COMPLIANCE_TRACKER.md or university-guidance/FORMAT_REQUIREMENTS.md

Use the existing template files as starting points.

If your project is a dissertation or thesis, you may also create DISSERTATION_BRIEF.md from DISSERTATION_BRIEF_TEMPLATE.md.

3. Configure Agent Memory

Update:

• AGENTS.md
• PROJECT_AGENT_PREFERENCES.md
• research-wiki/TASK_STATE.md

Keep private information minimal. Use placeholders until a source file confirms a fact.

4. Use Skills

Project skills live in:

.agents/skills/

Use the skills by name in agent prompts, or let agent-orchestration route tasks.

For formal writing, keep these files active:

• .agents/skills/cognitive-frameworks/SKILL.md
• .agents/skills/academic-integrity-preflight/SKILL.md
• .agents/skills/academic-self-review-loop/SKILL.md
• .agents/skills/authorial-voice-integrity/SKILL.md
• research-wiki/WRITING_QUALITY_RUBRIC.md
• research-wiki/AI_WRITING_AUTHORIAL_VOICE_POLICY.md
• research-wiki/DOCUMENT_PIPELINE.md

They help the agent package evidence status, check integrity, plan the argument, self-review the draft, revise authorial voice safely, and record thinking/writing/delivery checkpoints.

For knowledge-base work, start here:

• knowledge-base/self-growing/README.md
• knowledge-base/self-growing/growth-queue.md
• knowledge-base/self-growing/compiled-wiki/INDEX.md

Then run:

python scripts/kb_health_check.py
python scripts/build_agent_index.py --rebuild --summary
python scripts/local_retrieval_search.py --rebuild --query "source readiness"

Optional specialist skills:

• research-neural-network-figure for real AI/model architecture diagrams.
• research-nature-figure for high-quality research figure planning.
• research-nature-writing for article-style structure after evidence checks.

These skills are quality layers. They do not replace source checks, compliance checks, citation review, or delivery gates.

5. Keep Evidence Boundaries

Use labels:

• CONFIRMED
• LITERATURE-SUPPORTED
• CONTEXTUAL
• INFERENCE
• TO CONFIRM
• NEEDS VERIFICATION

6. Before Sharing

Run:

./scripts/privacy_check.sh
python scripts/run_skill_evals.py
python scripts/validate_agent_schemas.py
python -m unittest discover -s tests

Review the output before pushing to GitHub or sharing with friends.