Overview

Overview

Rapidly build a mental model of an unfamiliar codebase by analysing its commit history. Produces month-by-month graded summaries, a synopsis, and an honest overall assessment.

Best used when taking on a brownfield project — understand what was built, how well it was built, and where the risks are.

Workflow

Phase 1: Scope

git log --oneline --all

Determine:

• Date range (first to last commit)
• Total semantic commits (feat, fix, refactor, perf, docs prefixes)
• By default, exclude chore commits (release bumps, dependency updates, config tweaks). Include them if the user passes include:chore
• If an author-to-exclude argument was provided, exclude their commits
• Group remaining commits by month, show counts to user

Present the scope summary and ask user to confirm before proceeding.

Phase 2: Set up output directory

Derive the project name from the git remote (origin) or fall back to the current directory name.

# e.g. git@github.com:user/cool-app.git → cool-app
PROJECT_NAME=$(basename -s .git "$(git remote get-url origin 2>/dev/null)" 2>/dev/null || basename "$PWD")
OUTPUT_DIR="$HOME/overviews/$PROJECT_NAME"
mkdir -p "$OUTPUT_DIR"

All output goes to ~/overviews/{project-name}/ — never into the project repo itself.

Phase 3: Generate monthly documents

For each month with commits, create ~/overviews/{project-name}/YYYY-MM.md.

Use parallel subagents — one per month — to speed this up. Give each subagent the month range and repo path.

Subagent prompt template

Analyse all semantic commits in YYYY-MM for this repo. For each commit:
1. Read the actual diff (`git show <hash>`) — don't just restate the message
2. Write a 2-4 sentence explanation of what changed and why it matters
3. Grade the commit A-F with a score out of 100

Grading criteria:
- Code quality: clean, well-structured, appropriate abstraction level
- Commit hygiene: right-sized change, clear message, single concern
- Impact: does this move the product forward meaningfully

Be honest — accuracy over flattery.

Write the output to ~/overviews/{project-name}/YYYY-MM.md in this format:

# [Month Name] YYYY

**Theme:** [1-2 sentence summary of the month's focus]

---

### YYYY-MM-DD — commit message
**Hash:** `full-hash`
**Author:** developer name
**Grade:** B+ (82)
**Files changed:** N files (+X, -Y)

[2-4 sentence explanation based on the diff]

---

[repeat for each commit]

Include a header with month name, commit count, and a 1-2 line theme summary.
Exclude chore commits unless user opted in with include:chore.

Phase 4: Developer report cards

Create ~/overviews/{project-name}/DEVELOPERS.md. For each developer who contributed:

• Name and commit count
• Grade (A-F) with score out of 100
• Strengths (with specific commit examples)
• Weaknesses (with specific commit examples)
• Areas of the codebase they own (where most of their commits land)
• Summary: one paragraph characterising their work

Rank developers by volume, then by grade. This is the "who built what and how well" view.

Phase 5: Synopsis

After all monthly documents are written, create ~/overviews/{project-name}/SYNOPSIS.md:

• One section per month
• Month name and theme
• Grade (A-F) and score for the month overall
• 3-5 sentence summary of the most important things that happened
• ASCII quality trajectory chart at the bottom showing grade over time

Phase 6: Overall assessment

Create ~/overviews/{project-name}/OVERALL.md — a one-page honest assessment:

• What was built (brief)
• What was done well (with specific examples from the commits)
• What could be better (with specific examples)
• The arc of the project (how quality changed over time)
• A single overall grade A-F with score out of 100

Be honest. The point is to understand what you're walking into.

Key Rules

• Output goes to ~/overviews/{project-name}/ — never into the project repo
• Grading must be honest — accuracy over flattery
• Read actual diffs, not just commit messages
• chore commits excluded by default, included with include:chore argument
• Parallel subagents for months with many commits
• Focus on surfacing risks, patterns, and quality trends — this is due diligence