Claude Code Hooks & Subagent Scripts Library
You let Claude Code write code. But who's watching Claude?This library gives you 25 drop-in hooks and subagent configs that make Claude Code safer, faster, and observable — without changing how you work.Drop the files into ~/.claude/. Wire one settings template. Done.6 PreToolUse Hooks — run before Claude touches anything Secret scanner that blocks writes containing AWS keys, Stripe keys, private PEM blocks, and JWT tokens - Dangerous command guard that intercepts rm -rf, DROP TABLE, git push --force, terraform destroy - Branch protection that warns (or blocks) edits directly on main/master - Auto-linter that runs ESLint or Ruff before every file edit — blocks on errors - Test gate that requires your test suite to pass before Claude edits production files - Large file warning when a file exceeds 500 lines 5 PostToolUse Hooks — run after every write Auto-formatter: prettier, black, rustfmt, gofmt — runs automatically, nothing extra needed - Git auto-stage: git add after every successful edit — nothing goes unstaged - Smart test runner: finds and runs only the test file related to what Claude just changed - Token tracker: logs every tool call's input/output tokens and USD cost to a CSV - Desktop notifications when long-running Bash commands finish (macOS, Linux, Windows) 4 Stop Hooks — run when the session ends Session summary: Markdown file with files changed, recent commits, git diff stat, and session cost - Auto-commit: staged changes committed with an auto-generated message - Cost report: full breakdown of where your tokens went, per tool - Cleanup: removes pycache, .pyc files, and *.tmp files automatically 6 Subagent Configs (CLAUDE.md + settings.json pairs)Each one is a specialized Claude with a narrow role and hardened permissions: Reviewer — code review only, no edits, structured severity-rated output - Tester — writes tests for whatever you point it at, then runs them - Doc Writer — READMEs, API docs, inline comments — cannot touch source code - Security Scanner — OWASP Top 10 + more, runs npm audit / pip-audit / cargo audit - Refactor Bot — improves structure without changing behavior, runs tests to verify - Nested Orchestrator — coordinates the others in parallel or sequence (July 2026 nested agents feature) 4 Settings TemplatesPick one, copy it to ~/.claude/settings.json, and you're configured: settings-solo-dev.json — speed-optimized for personal projects - settings-team.json — conservative for shared repos - settings-open-source.json — strict write permissions for OSS maintainers - settings-yolo.json — maximum automation, minimum confirmation prompts Setup takes under 5 minutes. The included INSTALL.md has copy-paste commands for every step.All scripts are plain bash — readable, forkable, and adaptable to your workflow. No dependencies beyond Python 3.8+ (used for JSON parsing). Formatters and linters are auto-detected — if you don't have them installed, the hook silently skips.What you get: 15 bash hook scripts (pre-tool, post-tool, stop) - 6 subagent CLAUDE.md files - 6 subagent settings.json files - 4 settings templates - Full INSTALL.md setup guide - Example session log showing token-tracker output Tested on macOS and Linux. Windows notes in every script (WSL and Git Bash supported).$39 launch price — price goes to $49 after launch week.
Get it → isared.gumroad.com