Example: Full Runbook Run

End-to-end walkthrough of Junior reviewing a test repository with intentional bugs. Each step shows the exact data flowing through the runbook — useful for understanding what Junior collects, what the LLM receives, and what the output looks like.

The walkthrough uses three example prompts (security.md, logic.md, design.md). They are merged into a single system prompt and the review runs as one structured LLM call (see steps 3 and 4).

You can reproduce this by running:

cd /path/to/test-repo
OPENAI_API_KEY=sk-... junior run --harness pydantic \
  --prompt-file ./prompts/security.md \
  --prompt-file ./prompts/logic.md \
  --prompt-file ./prompts/design.md \
  --source branch -v

Steps

Step	File	Description
0	00_test_repo.md	Test repository setup and intentional bugs
1	01_collect.md	Phase 1: Collect — git diff, changed files, metadata
2	02_context_build.md	Context builder: the user message sent to AI
3	03_prompts.md	System prompts: how the prompts are merged into one system prompt
4	04_review.md	Phase 2: AI review — the structured ReviewResult returned by the LLM
5	05_publish.md	Phase 3: Formatted markdown output

Configuration Used

Setting	Value
Collector	local (no platform API)
Harness	pydantic (single structured call)
Publisher	local (stdout)
Provider	openai
Model	gpt-5.4-mini
Prompts	security.md, logic.md, design.md (merged into one system prompt)