Free client-side tool
Hooks Generator
Generate a valid Claude Code settings.json hooks block for lint gates, Stop checks, notifications, audit hooks, and secret blockers.
Open toolClaude Code engineering practice
Ship code with Claude, but keep the engineering loop intact.
A source-backed field manual for hooks, MCP, skills, subagents, CI, settings, permissions, and automation. Built for teams that want reviewable diffs, not AI theater.
Signature experience · synthetic
The Terminal
A labeled synthetic Claude Code session you can step through: version check, a/verify skill as used in Code, a read-only MCP fetch plus a Stop hook, then a shipping check whose output is still in the session. Practice reviewing the transcript the way Claude Logs teaches — claims versus evidence — on a loop that is supposed to ship.
This site does not pin a CLI version. Run claude --version locally. Skill authoring belongs on Claude Has Skills. Forensic specimens belong on the Black Box.
Four beats
What you step through.
- Version
claude --version— unpinned. Confirm your binary, then read the changelog. - Skill
/verifyloads a project SKILL.md. Authoring lives on Claude Has Skills. - Hook / MCPRead-only ticket fetch, then a Stop hook that refuses a premature done.
- Check
npm test -- hooksandgit diff --statbefore the summary.
Key facts
The site's operating model
- Claude Code work should move through context gathering, scoped edits, verification, review, and deploy gates.
- The Terminal at /terminal/ is a labeled synthetic session for practicing that loop. It is not a real transcript. Forensic log reading belongs on Claude Logs.
- As of September 2026, Claude Fable 5.1 (
claude-fable-5-1) defaults to High effort inside Claude Code and Medium on Claude.ai and Cowork, so budget cost and latency from the surface you actually run. - Claude Opus 5 (
claude-opus-5) is the general starting point, with a Fast mode at roughly 2.5x speed for 2x base price through Claude Code usage credits. Opus 4.8 is legacy. - Hooks enforce deterministic rules; MCP connects live external systems; skills and subagents manage reusable or isolated work.
- Free tools on this site generate starter config locally in the browser for hooks, permissions, and slash commands.
Defaults as of 2026-09-19
Check your model and effort defaults before you budget a session.
Model defaults are engineering configuration: they set session cost, how long an agent can run before it stalls, and how much latency a reviewer waits through. These are the current generally available options and the IDs to pin.
claude-fable-5-1Fable 5.1, shipped September 1, 2026. Most capable generally available model. High effort by default in Claude Code, Medium on Claude.ai and Cowork. $10/$50 per MTok, 1M context, cache reads at 2.5% of base input.claude-opus-5Opus 5, shipped July 24, 2026. The recommended starting point for most workloads. $5/$25 per MTok, default effort high. Fast mode runs ~2.5x faster at 2x base price via platform access or Claude Code usage credits.claude-sonnet-5Sonnet 5. $2/$10 per MTok for high-volume or routine agent work. The introductory rate became the standard rate; it did not rise on September 1, 2026.claude-haiku-4-5Haiku 4.5. $1/$5 per MTok, 200K context. Fastest and cheapest, suited to narrow mechanical subagent tasks rather than long-horizon repository work.
Free tools
Generate starter config, then review it like code.
The tools are client-side utilities for Claude Code settings and workflow files. They target high-intent setup questions without replacing team review.
Free client-side tool
Permissions Builder
Build a Claude Code permissions.allow, permissions.ask, and permissions.deny profile for a repo risk posture, tools, MCP servers, agents, and file boundaries.
Open toolFree client-side tool
Command Scaffolder
Generate a .claude/commands markdown file with frontmatter, argument hints, allowed tools, workflow instructions, and verification expectations.
Open toolManual
Start with the workflow, then choose the mechanism.
The pages below stay inside Claude Code engineering practice. Adjacent topics link to sibling network sites instead of duplicating them.
Start here
Claude Code Tutorial: Ship One Verified Change
A practical Claude Code tutorial for moving from first repository session to a verified, reviewable change without relying on vibes.
Operating modelClaude Code Best Practices for Engineering Teams
Source-backed Claude Code best practices for teams: explore first, keep context small, verify work, use deterministic gates, and review diffs.
Project instructionsCLAUDE.md and Settings: What Teams Should Actually Configure
A practical guide to CLAUDE.md, settings scopes, project instructions, and local overrides for Claude Code engineering teams.
Settings examplesClaude Code settings.json Examples for Teams
Practical Claude Code settings.json examples for permissions, hooks, slash commands, local overrides, and team-safe configuration.
Deterministic automationClaude Code Hooks: Deterministic Gates for Agentic Coding
A practical Claude Code hooks guide covering lifecycle events, stop gates, formatting, tests, security, and when hooks beat instructions.
External systemsClaude Code MCP: Connect Tools Without Copy-Paste Context
A source-backed guide to using Model Context Protocol with Claude Code for issue trackers, databases, monitoring, design tools, and internal APIs.
Lane discipline
What this site covers
- Claude Code tutorial and operating model.
- The Terminal: a labeled synthetic shipping-session theater.
- CI, GitHub Actions, and reviewable automation.
- Hooks, MCP, settings, skills, subagents, and Agent SDK choices.
- Model and effort defaults as they affect Claude Code cost, latency, and CI.
- Security and permission boundaries for engineering teams.
Adjacent sites
What this site links out for
- Claude WeeklyNews and weekly updates.
- Claude Has SkillsSkill directory and authoring examples.
- Claude MemoriesDeep memory workflows.
- Claude ContextContext-window strategy.
- Claude HelpsEveryday how-to, including Claude in Chrome.
Primary-source base
Built from official docs and repositories.
Core claims cite Anthropic's Claude Code docs, the official Claude Code Action repository, and Model Context Protocol documentation. Start withbest practices,the extension overview, andthe changelog.