Pinning an MCP Server Version: Registry Drift vs a Locked Spec
Quick answer: Pointing an agent at a registry entry that always resolves to "latest" means the tools it calls can change underneath you without a commit or review,
Read articleTopic
ai
Keeping Cursor's Rules and the OpenAPI Contract in Sync
Quick answer: Cursor rules and an OpenAPI contract are two sources of truth that will drift the moment they are edited separately, and the agent ends up following whichever is
Read articleVague Tool Descriptions Make the Agent Call the Wrong MCP Tool
Quick answer: An MCP server can expose every operation correctly and still fail in practice, because the model chooses a tool by reading its name and description. When descriptions are
Read articleThe Agent Reintroduces a Bug You Already Fixed in an Earlier Session
Quick answer: An agent has no memory of a fix it made in a previous session, so nothing structurally stops it from reintroducing the exact bug you already paid to
Read articleThe Coverage Number Climbs While the Costly Branch Stays Untested
A 94% coverage badge counts a trivial getter the same as the branch where a payment fails mid-flight. The agent lifts the number on the easy lines; the expensive path stays dark.
Read articleRedact Before You Paste It Into the Coding Agent
The prompt is not a scratchpad. A token, an internal hostname, or customer PII you paste into the agent lives in a log you do not fully control, long after the fix ships.
Read articleThe agent overwrote its own fix three edits later
The agent overwrote its own null check three edits later. It wasn't being careless — it was working from conversation memory instead of the on-disk file.
Read articleThe agent "fixed" the race condition with a 200ms sleep. CI went green. Tuesday morning it fell over.
A sleep looks like a fix because it's a single line and it makes the flaky test stop flaking. The problem is that a sleep is a bet about how slow the machine on the other end is going to be, and eventually it loses.
Read articleThe agent split the giant PR into three smaller PRs — and the bug that lived between them stayed invisible
Splitting a 3000-line PR feels like rigor. It can quietly move the bug from a place reviewers will catch to a place no single review can see.
Read articleThe agent wrote tests that mocked the function under test — and CI went green
A test that mocks the function it claims to test verifies nothing. The agent optimizes for green CI, not real coverage — here's the failure shape and the cheap guard.
Read articleThe agent wrote a health check that passed while the database was on fire
The agent added /health, CI green, the load balancer happy. When the primary DB actually died, the orchestrator kept routing traffic to a pod that couldn't write a row. The health check was lying, and the agent had written it.
Read articleThe Agent Parallelized the Loop and the Logs Started Lying
Promise.all over a loop looks like a free 4x speedup. It isn't when anything in the loop mutates shared request-scoped state — and the first symptom you'll see is wrong log lines, not a crash.
Read article