Implement an automated 'fix and commit' loop by configuring an AI stop hook. This hook runs quality checks on AI-generated code and automatically prompts the AI to fix any detected errors.
with Claire Vo
Using the Claude Agent SDK, write a script (e.g., in TypeScript) that will run automatically every time the AI agent finishes a task.
Add logic to the script to check for file changes. If files were modified by the AI, run a quality gate command like a linter or type checker (e.g., bun typecheck).
If the quality check fails, program the script to capture the error report and output a new JSON prompt instructing the AI to fix the specific errors. The script's console.log output is sent back to Claude.
{"prompt": "Please fix the TypeScript errors. Here is the report: ..."}If all checks pass, the script can trigger a subsequent action, such as instructing a background agent to generate a git commit message and commit the successful changes.
Link your script to the AI tool by adding a configuration to its settings file (e.g., settings.local.json for Claude Code), specifying the command to run the hook.
{
"claude.hooks.stop": [
{
"command": "bun run claude-hooks/index.ts"
}
]
}
Speed up your development workflow by creating shortcuts for frequent AI commands using shell aliases and building dedicated command-line tools to script complex, repetitive AI tasks.
Boost your AI coding assistant's performance by preloading your application's architecture using text-based Mermaid diagrams. This gives the AI deep context, resulting in faster, more accurate code-related answers.
Improve your AI's performance by moving from a single, massive context file to a modular library of small, focused markdown files. This system allows the AI to intelligently load only the most relevant information for each task, enabling you to use simple, 'lazy' prompts and get better results.
We've made ChatPRD affordable so everyone from engineers to founders to Chief Product Officers can benefit from an AI PM.