Forem: Max

Build an AI Coding Bot That Fixes GitHub Issues While You Sleep

Max — Mon, 06 Apr 2026 09:10:58 +0000

Every morning, you open GitHub to a list of issues. Bug reports, small features, refactoring tasks. You triage, assign, context-switch, and by 3 PM you have written maybe 200 lines of actual code.

What if half those issues were already fixed — with tests passing and a PR waiting for your review?

That is exactly what an async SWE coding bot does. It monitors your repo, picks up labeled issues, writes code, runs tests, and submits pull requests. You stay in the loop as the reviewer, not the typist.

The Problem: Context Switching Is Killing Your Output

Studies show developers lose 23 minutes every time they context-switch. If you handle 8 issues a day, that is over 3 hours lost just getting back into flow. The actual coding might take 30 minutes per issue — but the overhead makes it feel like a full day.

The bottleneck is not writing code. It is the decision-making loop: read the issue, understand the context, find the relevant files, write the fix, test it, push it. Most of that is mechanical.

The Solution: An Autonomous Coding Agent in Your CI Pipeline

Here is how to build a bot that handles the mechanical parts while you focus on architecture and product decisions.

Step 1: Set Up Issue Labeling

Create a label in your GitHub repo called ai-fix. This is your trigger. When you (or your team) triages an issue and decides it is suitable for automated fixing, slap the label on it.

Good candidates for ai-fix:

Bug fixes with clear reproduction steps
"Change X to Y" refactoring tasks
Adding tests for existing functionality
Documentation updates
Simple feature additions with clear specs

Step 2: Configure the Agent Workflow

The bot watches for the ai-fix label via GitHub webhooks. When triggered, it:

Reads the issue — title, body, comments, linked PRs
Clones the repo into an isolated environment
Analyzes the codebase — finds relevant files using semantic search
Writes the fix — generates code changes based on the issue description
Runs your test suite — ensures nothing breaks
Submits a PR — with a clear description linking back to the issue

Step 3: Add Human Review Guardrails

This is critical. The bot should never merge its own PRs. Every change goes through your normal review process:

PR description includes the reasoning and approach taken
CI runs automatically (linting, tests, type checks)
A human reviewer approves or requests changes
The bot can iterate based on review comments

This keeps the quality bar high while eliminating the initial coding effort.

Step 4: Scale With Parallel Execution

Once the basic loop works, the real power kicks in. The bot can work on multiple issues simultaneously — each in its own branch and isolated environment. While you review PR #1, the bot is already working on issues #2 through #5.

A team of 4 developers with this setup can effectively output the work of 6-7, without burning out or hiring.

Real Results

Teams using this pattern report:

40-60% reduction in time-to-fix for routine issues
3x more PRs merged per week
Developers spending more time on design and architecture (the high-value work)
Faster onboarding — new team members see AI-generated PRs as learning examples

Getting Started

The full implementation guide walks you through setting up the GitHub webhook, configuring the AI agent with proper codebase context, writing the test-and-submit pipeline, and adding safety rails.

👉 Read the complete guide on Terminal Skills

Related Resources

Build a Multi-Agent Claude Code Team — run multiple AI agents in parallel for even faster development
Build Parallel AI Agent Workflows with Git Worktrees — use Worktrunk to manage concurrent agent branches without conflicts

The shift is not about replacing developers. It is about eliminating the repetitive 80% so you can focus on the creative 20% that actually moves your product forward. Set up the bot, label the easy issues, and wake up to a stack of PRs ready for review.

Run 5 Claude Code Agents in Parallel and Ship a SaaS MVP in 2 Days

Max — Fri, 03 Apr 2026 09:06:38 +0000

Most developers write code one feature at a time, in one terminal window, waiting for Claude to finish before starting the next task. That is the bottleneck. What if you ran 5 Claude Code agents in parallel — each owning a different slice of the codebase — and shipped a full-stack SaaS MVP in 2 days instead of 2 weeks?

That is exactly the workflow described in this use-case on terminalskills.io. Here is how it works and why it changes everything.

The Problem with Single-Agent Coding

When you use Claude Code or Cursor in a single session, you are essentially pair-programming — one human, one AI, one thread. That works well for focused, greenfield code. But real projects have parallel workstreams:

Frontend components being built while the API routes are still being designed
Database schema evolving while the auth layer is being wired up
Tests being written for features not yet finished

A single agent cannot do all this at once without getting confused by context switches, merge conflicts, and competing goals. The solution is to stop thinking of AI coding as a single-agent task and start treating it like a small engineering team.

The Multi-Agent Setup

The approach uses 5 Claude Code sessions running simultaneously, each given a specific domain:

Agent 1 — Architecture — owns /infrastructure, docker-compose.yml, environment config
Agent 2 — Backend/API — owns /server, route handlers, business logic
Agent 3 — Database — owns /migrations, schema design, seed data
Agent 4 — Frontend — owns /client, components, state management
Agent 5 — Tests & CI — owns /tests, GitHub Actions pipelines, coverage config

Each agent gets a dedicated CLAUDE.md in its directory with its specific role, the interfaces it must respect (types, API contracts), and strict instructions not to touch other agents directories.

How Coordination Works

The key insight is that agents communicate through shared contracts, not through each other. You define the API schema upfront (OpenAPI spec or TypeScript interfaces) and each agent programs to that contract independently.

A shared CONTRACTS.md file in the repo root defines:

API endpoint signatures
Database table schemas
Frontend component prop types
Event names for the message bus

When an agent discovers a contract needs to change, they flag it in a CONFLICTS.md file. The human reviews conflicts once per hour, resolves them, and updates CONTRACTS.md. Agents then pick up the updated contracts on their next task.

This is basically how a real engineering team works — except the standup is async and takes 10 minutes instead of 30.

What a 2-Day Sprint Looks Like

Day 1:

Hour 0–1: You define the MVP scope, write CONTRACTS.md, create 5 agent workspaces
Hours 1–8: Agents run in parallel. You check CONFLICTS.md every 90 minutes.
End of Day 1: Backend API is functional, database schema is finalized, frontend shell is rendering

Day 2:

Hours 0–4: Agents wire everything together. Agent 5 runs tests continuously.
Hours 4–6: You do integration review, merge to main, fix the 3–5 edge cases that always slip through
Hours 6–8: Deploy, smoke test, done

The result is a working SaaS MVP — auth, billing hooks, core feature loop, basic CI — in 48 hours of wall-clock time. Not 2 weeks.

Why This Works (And When It Breaks)

It works because:

Large language models are stateless by default, so running 5 of them in parallel is trivially cheap
Separation of concerns is a solved problem — agents stay in their lanes if you define lanes clearly
Context windows stay small per agent, which means better output quality

It breaks when:

Contracts are undefined or vague — agents will make incompatible assumptions
One agent modifies shared infrastructure unexpectedly — always lock infrastructure to one owner
You go more than 3 hours without a conflict review — technical debt compounds fast

The Numbers

Teams using this workflow consistently report:

5–8x reduction in time-to-MVP vs. single-agent or solo coding
60–70% of code passing review on first pass (agents with tight context produce cleaner output)
2–3 conflict resolutions per day on average — far fewer than you would expect

Get Started

The full workflow, including the CLAUDE.md templates for each agent role, the CONTRACTS.md starter format, and the git branch strategy, is documented in the complete guide:

Read full guide → Build a Multi-Agent Claude Code Team on terminalskills.io

Related use-cases worth reading alongside this one:

Build Parallel AI Agent Workflow with Git Worktrees — run agents on isolated git branches with zero merge conflicts
Optimize AI Coding Agent for Production Team Use — standardize Claude Code across your whole team with shared CLAUDE.md templates

The shift from single-agent to multi-agent coding is the same leap as going from solo freelancer to a coordinated team. You are not just going faster — you are fundamentally changing what is possible in a given day. Stop waiting for one agent to finish. Run the team.

Build an AI Short Video Generator: 50+ Faceless Videos Per Day on Autopilot

Max — Wed, 01 Apr 2026 09:01:18 +0000

Every day, faceless YouTube channels and TikTok accounts publish dozens of short videos and quietly collect ad revenue, affiliate clicks, and sponsorships — all without showing a face, hiring a team, or touching a video editor.

The secret? A fully automated AI pipeline that handles everything from topic research to upload.

In this guide, I will walk you through how to build exactly that: a system that generates 50+ short videos per day using AI agents, text-to-speech, automated editing, and direct platform uploads.

The Problem With Manual Short-Form Content

If you have ever tried to consistently post on YouTube Shorts or TikTok, you know the grind. Coming up with topics takes time. Writing scripts takes time. Recording, editing, adding captions, uploading — it stacks up fast.

Most creators burn out within weeks. The few who last either outsource to a team (expensive) or find a way to automate.

The good news: in 2026, the full stack exists to automate almost every step.

The Architecture: 4 Stages

Here is the high-level pipeline:

Topic Research → Script Generation → Video Assembly → Upload & Schedule

Each stage can run independently, which means you can start simple and layer on automation as you go.

Stage 1: Topic Research

Use an AI agent to query trending topics from Google Trends, Reddit, and YouTube search autocomplete. Filter by search volume, competition, and monetization potential.

# Example: agent scans Reddit + Trends, outputs top 20 topic ideas
node scripts/topic-agent.js --niche "personal finance" --count 20 > topics.json

The output is a JSON file with titles, hooks, estimated search volume, and competitor gap score.

Stage 2: Script + Voiceover Generation

For each topic, an LLM generates a 45–60 second script optimized for retention (hook in the first 3 seconds, pattern interrupts every 10 seconds, CTA at the end). Then ElevenLabs or a local TTS model renders the audio.

const script = await generateScript(topic, { duration: 55, style: "educational" });
const audio = await textToSpeech(script, { voice: "energetic-male", speed: 1.1 });

Stage 3: Video Assembly

This is where ffmpeg and a stock footage API (Pexels, Pixabay, or Storyblocks) come together. The agent:

Pulls B-roll clips matching keywords from the script
Overlays auto-generated captions (word-by-word, TikTok style)
Adds background music at 20% volume
Renders a 1080x1920 vertical video

ffmpeg -i background.mp4 -i voiceover.mp3 \
  -vf "subtitles=captions.srt:force_style=FontSize=22" \
  -c:v libx264 -preset fast output.mp4

Stage 4: Upload & Schedule

The final agent logs into YouTube and TikTok via their APIs, uploads the video with an AI-generated title, description, tags, and thumbnail, and schedules it for peak engagement hours.

await youtubeUpload({ file: "output.mp4", title, description, tags, scheduledAt });
await tiktokUpload({ file: "output.mp4", caption, hashtags });

What You Need to Get Started

Node.js or Python for scripting the agents
ElevenLabs API (or a local model like Coqui TTS) for voiceovers
Pexels API (free) for stock footage
FFmpeg for video assembly
YouTube Data API + TikTok Content Posting API for uploads

You can run the whole pipeline on a $6/month VPS. At 50 videos/day, you are looking at roughly $0.05 per video in API costs.

Real Results

Channels running this type of pipeline typically see:

Month 1: 50–200 subscribers, channel finding its niche
Month 3: YouTube monetization threshold hit (1,000 subs + 4,000 watch hours)
Month 6: $300–$1,500/month from ads + affiliate links embedded in descriptions

The key is consistency and niche selection. Broad niches are crowded. Specific niches (AI tools for accountants, minimalist productivity, beginner investing) compound faster.

Read the Full Guide

For the complete step-by-step implementation — including agent code, FFmpeg templates, scheduling logic, and channel growth strategies:

👉 Build an AI Short Video Generator — Terminal Skills

Build an AI Content Monetization Pipeline — turn content into multiple revenue streams automatically
Build an AI Agent That Clones Any Website — reverse-engineer any site and rebuild it in under an hour

What niche would you target first with an automated short-form video channel? Finance, tech tutorials, something else? Drop it in the comments — curious to see what the dev community would build.

Build a Cold Email Engine That Writes Hyper-Personalized Emails with AI

Max — Tue, 31 Mar 2026 09:53:51 +0000

Every B2B founder knows cold email works. But writing 1000 personalized emails? That takes weeks.

What if an AI could research each prospect and write a unique email -- not just "Hi {name}" but genuinely personalized based on their company, role, and recent activity?

The Problem

Generic cold emails get 2% reply rates. Personalized ones get 15-25%. But manual personalization at scale is impossible.

The Solution: AI Research + Personalization Pipeline

Step 1: Prospect List

Start with a CSV: name, company, LinkedIn URL, website.

Step 2: AI Research

For each prospect, the agent scrapes their company website with Firecrawl, reads their LinkedIn summary, and identifies what they do, recent news, tech stack, and pain points.

Step 3: Personalized Email

Claude writes a unique opener for each prospect based on the research. Not template-y. Not generic. Actually relevant.

Step 4: Send + Track

Send via Resend with custom domain. A/B test subject lines. Track opens and replies. Auto-pause when prospect replies.

Results

1000 personalized emails in 4 hours (vs 2 weeks manually)
18% reply rate (vs 2% generic)
Cost: about 15 dollars in API calls

Read the Full Guide

Complete code, email templates, and Resend setup:

Build Cold Email Personalization with AI

Related Guides

All 900+ guides on Terminal Skills.

Whats your cold email reply rate? Has AI improved it? Share below!

Ship a Micro-SaaS in 48 Hours: A Weekend Playbook with AI

Max — Tue, 31 Mar 2026 09:41:32 +0000

You have a SaaS idea. You've been thinking about it for months. But building it feels like a 3-month project.

What if you could go from idea to paying customers in 48 hours?

Here's the exact playbook — based on a real use-case where a developer shipped "ScreenInvoice" (screenshot → invoice PDF) in one weekend.

Day 1, Morning: Validate in 2 Hours

Before writing a single line of code:

Score your idea on 10 factors (urgency, market size, pricing potential, etc.). If it scores below 50/100, pick a different idea.

Shadow test: Spin up a landing page with Vercel + Next.js. Run $50 in Google Ads. If you get 20+ signups in 24 hours, you have demand.

Day 1, Afternoon: Build the Core (4 Hours)

Only build the ONE thing that makes your product valuable. Auth with Clerk (5 min setup), payments with Stripe Checkout (copy-paste), database with Prisma + Neon Postgres.

Day 1, Evening: Landing Page

Your page needs 4 things: headline, demo GIF, pricing ($19/month), and CTA button → Stripe Checkout.

Day 2: Polish, Deploy, Launch

Deploy to Vercel (free). Post on Twitter/X, Hacker News, Reddit, IndieHackers.

The math: 16 hours of work, $0 cost, goal of 3 paying customers = $57 MRR = $684/year from a weekend.

Read the Full Guide

👉 Build a Micro-SaaS in a Weekend — Full Guide

All guides on Terminal Skills — 900+ AI coding skills and use-cases.

What micro-SaaS would YOU build in a weekend? Drop your idea below 👇

7 AI Skills That Will 10x Your Coding Speed in 2026

Max — Tue, 31 Mar 2026 09:33:55 +0000

AI coding agents are everywhere now — Claude Code, Cursor, Codex, Gemini CLI. But most developers use them like fancy autocomplete. The real power comes from skills — structured instructions that turn AI agents into specialized experts.

Here are 7 skills that dramatically changed how I code:

1. TDD (Test-Driven Development with AI)

Instead of asking AI to "write code", give it the TDD skill and watch it:

Write failing test first
Implement minimal code to pass
Refactor while keeping tests green
Repeat for each feature slice

The AI actually writes better code when constrained by tests. No more "looks good but doesn't work" output.

Try it: TDD Skill

2. Vibe Coding

Structure your project so AI agents can navigate it efficiently:

Clear CLAUDE.md / AGENTS.md with project context
Module boundaries that fit in context windows
Spec-first workflow: describe what you want → AI implements

Try it: Vibe Coding Skill

3. MCP Server Builder

Build custom tools that your AI agent can use:

Database queries via MCP
Internal API access
Custom data transformations

Your agent stops being a text-only assistant and becomes a tool-wielding power user.

Try it: MCP Server Skill

4. Prompt Engineering

Most developers don't know about:

Chain-of-thought for complex reasoning
XML tags for structured prompts
Few-shot examples for consistent output format
ReAct pattern for tool-using agents

Understanding these patterns makes your AI 3x more reliable.

Try it: Prompt Engineering Skill

5. Claude Code HUD

Monitor what your agent is actually doing:

Context window usage (don't hit the limit mid-task)
Active tool calls
Sub-agent progress
Cost tracking

Try it: Claude HUD Skill

6. AI Code Reviewer

Set up a GitHub Action that reviews every PR with AI:

Catches bugs before human reviewers
Flags security issues
Posts inline comments with suggestions
Configurable rules per language

Try it: AI Code Reviewer Use-Case

7. Worktrunk (Parallel Agents)

Run 5 Claude Code agents simultaneously:

Each works in its own git worktree
Different branch, different feature
Merge when done
2-week sprint → 2 days

Try it: Worktrunk Skill

Where to Find More

All these skills (and 900+ more) are available on Terminal Skills — an open-source marketplace of AI coding skills that work with Claude Code, Cursor, Codex, Gemini CLI, and more.

Install via CLI:

npm i -g terminal-skills
terminal-skills search "tdd"
terminal-skills install tdd

Or browse the full catalog.

Which AI skill would you try first? Share your favorites in the comments 👇

Build an AI Agent That Clones Any Website in Under an Hour

Max — Tue, 31 Mar 2026 09:33:28 +0000

Ever looked at a competitor's landing page and wished you could rebuild it in your own stack? With AI coding agents, you can now clone any website as a pixel-perfect Next.js app — in under an hour.

The Problem

You need a professional landing page. You love how Linear.app looks. But hiring a designer is $5k and 3 weeks. What if an AI agent could reverse-engineer the design, extract every color/font/spacing token, and rebuild it as clean React components?

How It Works

The workflow uses a multi-phase pipeline:

Reconnaissance — Screenshots at every breakpoint, extract design tokens (colors, fonts, spacing), sweep all interactions (hover, scroll, click)
Foundation — Download fonts, images, videos. Set up Tailwind config with extracted tokens
Component Specs — Write detailed specs for each section with exact getComputedStyle() values
Parallel Build — Dispatch builder agents in git worktrees, one per section
Assembly & QA — Merge worktrees, visual diff against original

Quick Start

# Clone the template
git clone https://github.com/JCodesMore/ai-website-cloner-template.git my-clone
cd my-clone && npm install

# Start Claude Code with Chrome MCP
claude --chrome

# Run the skill
/clone-website https://linear.app

The agent takes over from here — screenshots, extracts tokens, writes specs, builds components in parallel.

What Gets Extracted

Colors — Every hex/oklch value mapped to CSS custom properties
Typography — Font families, sizes, weights, line heights
Spacing — Margins, paddings, gaps as design tokens
Assets — Images, videos, SVGs downloaded and optimized
Interactions — Hover states, animations, scroll behaviors

The Result

After ~30-45 minutes, you have a fully functional Next.js app:

React 19 + Next.js App Router
Tailwind CSS v4 with extracted design tokens
shadcn/ui components
Responsive at all breakpoints
All animations and interactions preserved

Customize and Ship

The clone is your starting point. Swap in your copy, adjust colors to your brand, modify CTAs, and deploy to Vercel.

Try It Yourself

The clone-website skill is available on Terminal Skills — a marketplace of 900+ AI coding skills for Claude Code, Cursor, Codex, and more.

Check out the full use-case guide with step-by-step instructions.

What website would you clone first? Drop a comment below 👇

Forem: Max

Build an AI Coding Bot That Fixes GitHub Issues While You Sleep

The Problem: Context Switching Is Killing Your Output

The Solution: An Autonomous Coding Agent in Your CI Pipeline

Step 1: Set Up Issue Labeling

Step 2: Configure the Agent Workflow

Step 3: Add Human Review Guardrails

Step 4: Scale With Parallel Execution

Real Results

Getting Started

Related Resources

Run 5 Claude Code Agents in Parallel and Ship a SaaS MVP in 2 Days

The Problem with Single-Agent Coding

The Multi-Agent Setup

How Coordination Works

What a 2-Day Sprint Looks Like

Why This Works (And When It Breaks)

The Numbers

Get Started

Build an AI Short Video Generator: 50+ Faceless Videos Per Day on Autopilot

The Problem With Manual Short-Form Content

The Architecture: 4 Stages

Stage 1: Topic Research

Stage 2: Script + Voiceover Generation

Stage 3: Video Assembly

Stage 4: Upload & Schedule

What You Need to Get Started

Real Results

Read the Full Guide

You Might Also Like

Build a Cold Email Engine That Writes Hyper-Personalized Emails with AI

The Problem

The Solution: AI Research + Personalization Pipeline

Step 1: Prospect List

Step 2: AI Research

Step 3: Personalized Email

Step 4: Send + Track

Results

Read the Full Guide

Related Guides

Ship a Micro-SaaS in 48 Hours: A Weekend Playbook with AI

Day 1, Morning: Validate in 2 Hours

Day 1, Afternoon: Build the Core (4 Hours)

Day 1, Evening: Landing Page

Day 2: Polish, Deploy, Launch

Read the Full Guide

You Might Also Like

7 AI Skills That Will 10x Your Coding Speed in 2026

1. TDD (Test-Driven Development with AI)

2. Vibe Coding

3. MCP Server Builder

4. Prompt Engineering

5. Claude Code HUD

6. AI Code Reviewer

7. Worktrunk (Parallel Agents)

Where to Find More

Build an AI Agent That Clones Any Website in Under an Hour

The Problem

How It Works

Quick Start

What Gets Extracted

The Result

Customize and Ship

Try It Yourself