5 Agentic Coding Tips & Tricks

Introduction

Agentic coding only feels “smart” when it ships correct diffs, passes tests, and leaves a paper trail you can trust. The fastest way to get there is to stop asking an agent to “build a feature” and start giving it a workflow it cannot escape.

Why This Matters

Idealized agentic coding promises fully automated feature development, but the technical reality is that current agents often produce broad, untested refactors due to a lack of codebase understanding. This can lead to significant regression risk and wasted developer time—a single flawed agentic change can require hours or days of debugging and rollback.

Key Insights

• Diff budget enforcement: Limiting the size of code changes per iteration to 120 lines reduces complexity and improves reviewability.
• Executable tests as contracts: Translating requirements into failing tests provides agents with a clear, objective target and validates functionality.
• Run recipes for reproducibility: Requiring a RUN.md file with exact commands and environment details enables team members to replicate agent-generated changes.

Working Example

from pathlib import Path

INCLUDE_EXT = {".py", ".ts", ".tsx", ".go", ".java", ".rs"}
SKIP_DIRS = {"node_modules", ".git", "dist", "build", "__pycache__"}
root = Path(__file__).resolve().parents[1]
lines = []
for p in sorted(root.rglob("*")):
    if any(part in SKIP_DIRS for part in p.parts):
        continue
    if p.is_file() and p.suffix in INCLUDE_EXT:
        rel = p.relative_to(root)
        lines.append(str(rel))
print("\n".join(lines[:600]))

Practical Applications

• Stripe: Uses agentic coding to automate repetitive tasks like updating API documentation and generating boilerplate code.
• Pitfall: Relying on agents for large-scale refactoring without a clear understanding of the codebase can lead to unintended consequences and increased technical debt.

References:

• https://machinelearningmastery.com/5-agentic-coding-tips-tricks/