Three checks before I merge code an AI agent wrote
The agent says tests pass and the PR looks clean. That's the starting line, not the finish. Three cheap checks catch the failures that matter.
Read articleThe Powerduck Blog
API engineering, from the spec out
Product updates, engineering notes, and guides for designing, debugging, testing, mocking, and serving MCP tools from one OpenAPI spec.
Browse the guidesYour Agent Renamed the Function — It Just Missed the Caller That Referenced It by String
Static rename catches import statements. It does not catch the cron job, the route table, the config dict, or the dynamic dispatcher that still points at the old name. The build passes, the thing breaks at runtime.
Read articleYour Agent Wrapped the Whole Function in try/catch and Called It Error Handling
Paste a stack trace, get a try/catch around the whole function. The error still happens — now it returns undefined and the caller has no idea. Wrapping a call is not handling a failure.
Read articleYour Agent Writes the PR Description, but It Only Tells You What the Code Does
An agent-generated PR description reads the diff and narrates it back. It never explains why the change was needed, what could break, or what the author actually worried about. That last part is the whole point of review.
Read articleYour Agent Keeps Asking the Same Question It Already Answered Last Session
The agent from yesterday looked productive. The agent from today has forgotten every decision you made. Here's the cheap handoff that stops the reset.
Read articleYour Agent Will Refactor a Working Function and Quietly Drop the Edge Case
The diff looks cleaner. The tests pass. The function the original author spent three years defending is now shorter — and silently wrong for the input nobody remembers.
Read articleYour Agent Breaks the Dockerfile Because It Has Never Seen CI Run
Agents edit Dockerfiles and CI YAML like application code. They miss layer cache order, build-arg scoping, and the gap between your laptop and the runner.
Read articleLong agent sessions rot. Start a fresh one instead.
A 4-hour agent conversation starts re-suggesting things it already tried. The fix isn't more prompting. It's a checkpoint and a clean session.
Read articleYour AGENTS.md Should Survive a Tool Switch
The real cost of moving between Cursor, Cline, and Claude Code isn't the editor. It's re-explaining the same repo to each one from scratch. A short, tool-agnostic instructions file is the cheapest context tax you can cut.
Read articleEvery Agent Task Should End With a Command You Can Run
An agent's claim of 'done' is worthless without a one-line command that proves it. The cheap discipline: no task closes until the agent leaves you something you can paste into a terminal.
Read articleStop Pasting Your Whole Repo Into the Agent
Dumping every file into the prompt doesn't give the agent more signal — it buries the one file it actually needs. The cheap fix is a file-level context manifest, not a bigger context window.
Read articleYour Agent Is Fixing Flaky Tests by Quietly Weakening Them
When a coding agent hits a flaky test at 2am, it doesn't stabilize the test — it lowers the assertion until the suite goes green. Here's what that actually costs you, and the cheap gate that catches it.
Read article