Notes from shipping with Claude Code agents

What my agents get wrong, what my review catches, and the numbers from my own logs. I'm a senior full-stack engineer, 10+ years in production, and I review every PR my agents write.

Featured

The Shell-Injection Bugs My Agents Wrote, and the 3 I Missed

Agents wrote 7 shell commands built from strings into 30 PRs on my own repo. Review fixed 4 and missed 3. The diffs, the fixes and the check I run now.

AISecurity

July 9, 20268 min read

CLAUDE.md Examples: What I Cut and What Became a Hook

I cut my Claude Code rules from 10,667 words to 3,382. The rules that held became hooks. Real CLAUDE.md examples, and the escape hatch that undid a gate.

Claude CodeCLAUDE.md

October 17, 20268 min read

Claude Code Usage: 463 Sessions, 45B Tokens, Counted

66 days of my Claude Code usage, counted: cache writes were 2.1% of 45B tokens and 42% of the list-price dollars. Sessions, prompts, subagents and limits.

Claude CodeAI

October 10, 20269 min read

Where the AI Speed-Up Goes: Into the Review Queue

How much faster is AI coding? DX found under 8% more PRs in a sample from 400 companies. My agent PRs merged a median 44 minutes after opening.

AIDeveloper productivity

July 9, 20268 min read

How to Review AI-Generated Code: What 30 Agent PRs Needed

25 of 30 agent PRs needed a fix before merge. I counted all 79 fixes: duplicated code led, security came third, and review still missed 5 security bugs.

AICode review

July 9, 20268 min read

Notes from shipping with Claude Code agents

The Shell-Injection Bugs My Agents Wrote, and the 3 I Missed

CLAUDE.md Examples: What I Cut and What Became a Hook

Claude Code Usage: 463 Sessions, 45B Tokens, Counted

Where the AI Speed-Up Goes: Into the Review Queue

How to Review AI-Generated Code: What 30 Agent PRs Needed

Have a backlog? Book a 30‑minute scoping call.