Claude Code PR automation: babysit to green
Set up Claude Code PR automation that takes pull requests from red to green: diagnose CI failures, answer review threads, and escalate on cue.
articles
Technical writing on building, securing and operating systems with Claude.
20 articles published
All articles
Scoping Claude Code in monorepos
Scope Claude Code to the slice of your monorepo a task touches: layered CLAUDE.md files, path-scoped rules, sibling grants, and cheaper search.
CLAUDE.md best practices: contract not wiki
Turn CLAUDE.md best practices into a contract: few verifiable clauses, path-scoped rules, hooks for hard limits, and a repeatable adherence audit.
Claude Code context window management
Treat Claude Code context window management as a performance budget: profile with /context, cut MCP overhead, and spend tokens on real work.
When to interrupt Claude Code — or reset it
Learn when to interrupt Claude Code, when to rewind, and when to reset. A cost-ordered framework for corrections that compound with context depth.
Plan mode vs subagents: pick the right rung
Decide when to use plan mode vs subagents in Claude Code with an escalation ladder that prices every rung, from hand edits to headless fan-out.
Incident Response Runbooks with Claude Code
Build structured incident response runbooks as Claude Code skills that check service health, gather logs, and suggest root causes.
Secure API Middleware with Claude Code
Build JWT auth, rate limiting, input sanitization, and CORS middleware using security-first Claude Code prompts mapped to OWASP API risks.
Claude Code extension points: the matrix
Choose the right Claude Code extension point with one question: does the harness execute it, or does the model interpret it? Includes three migrations.
Claude Code headless mode in CI and cron
Run Claude Code headless in CI pipelines and cron jobs with JSON output, deny-by-default permissions, and bounded cost. Includes three full recipes.
Claude Code hooks as guardrails
Turn advisory CLAUDE.md rules into Claude Code hooks that block dangerous commands, verify every edit, and reload state after compaction.
Claude Code OpenTelemetry metrics, decoded
Turn Claude Code OpenTelemetry metrics and local JSONL transcripts into cost-per-task, interrupt-rate, and permission-friction data you act on.
Claude Code git worktrees in parallel
Run parallel Claude Code sessions in git worktrees: split work by ownership, freeze contracts first, and merge branches in dependency order.
Claude Code plan mode: plans worth a yes
Interrogate Claude Code plan mode output before approving. Covers plan anatomy, iteration prompts, a caught-assumption case, and when to skip planning.
Review AI-generated code in layers
Stop rubber-stamping giant diffs. Review AI-generated code with five layers: plans, tests, staged commits, adversarial sessions, and human eyes.
Claude Code prompt templates: spec-first
Copy three Claude Code prompt templates for bug fixes, features, and refactors, plus the four spec slots that let tasks land on the first pass.
Claude Code subagent context isolation
Use Claude Code subagents as context firewalls. Three patterns that keep test output, search noise, and review bias out of your main thread.
Claude Code auto-compact: survive the cliff
Learn what Claude Code auto-compact actually drops, why sessions degrade after compaction, and the disk-first workflow that makes resets free.
Tool use with Claude, explained
Understand the single API contract behind every Claude tool call — and how Claude Code's built-in tools, permissions, and agentic loop all connect to the same primitive.
The Claude Code verification loop ladder
Stop being your agent's test runner. Build Claude Code verification loops with TDD harnesses, Stop hooks, /goal, and fresh-context review.
Nothing in this pillar yet.