Python Langgraph Workflow

LangGraph Python Local CI & Git Push Workflow

This skill hooks into the project's native uv and Makefile lifecycle tasks, ensuring all LangGraph agents, validation suites, and formatting checks pass cleanly before committing changes.

🛠️ Environment Constraints

• Package Manager: uv
• Python Runtime Wrapper: uv run
• Task Runner: make

🏃‍♂️ Step 1: Pre-Commit Quality Verification

Run the unified check command. This will execute formatting checks, Ruff/Pylint linting, Mypy type-checking, Pytest suites, and codespell in sequence. Stop immediately if this command fails.

make check

Optional Specialized Targets

If a specific sub-task in the check fails, use these isolated commands to debug or fix:

• Auto-format Code: make format
• Run Unit Tests: make test
• Visualize LangGraph State Machine: Generate and inspect the compiled graph image:

make graph

🔍 Step 2: Stage & Diff Analysis

Once the codebase passes make check, audit the workspace modifications before committing.

1. Check Workspace Status:

git status

2. Stage Changes:

git add .

3. Review Staged Changes: Run the following command to review the exact lines modified in the graphs, prompts, or tests:

git diff --cached

• Agent Instruction: Read through the staged diff carefully. Note changes to LangGraph nodes, conditional edges, state schemas, or system prompts to write a precise, technical commit message.

🔀 Step 3: Git Commit & Push

Only proceed to this step after all local verification targets have passed and the staged changes have been audited.

1. Commit with Context:
   • Craft a concise, clear commit message using the imperative mood (e.g., feat: add conditional router edge to conversational agent graph).
   • Base the message directly on the specific logical changes found during the git diff --cached step.

git commit -m "<descriptive-message-from-diff>"

2. Push to Remote:

git push origin HEAD