@AGENTS.md
Rankings, higher = better. Cost reflects what I actually pay (OpenAI has really generous limits), not list price. Intelligence is how hard a problem you can hand the model unsupervised. Taste covers UI/UX, code quality, API design, and copy.
| model | cost | intelligence | taste |
|---|---|---|---|
| gpt-5.5 | 9 | 8 | 5 |
| sonnet-5 | 5 | 5 | 7 |
| <?xml version="1.0" encoding="UTF-8"?> | |
| <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> | |
| <plist version="1.0"> | |
| <dict> | |
| <key>Label</key> | |
| <string>com.clawbot.llama-server</string> | |
| <key>ProgramArguments</key> | |
| <array> | |
| <string>/opt/homebrew/bin/llama-server</string> |
Add those files to bin folder.
Usage:
bin/worktree-setup [task_id] - Sets up new worktree with isolated test DBbin/worktree-teardown - Drops test DB and prints removal command
| #!/bin/bash | |
| # Read JSON input | |
| input=$(cat) | |
| # Extract session data | |
| model=$(echo "$input" | jq -r '.model.display_name') | |
| limit=$(echo "$input" | jq -r '.context_window.context_window_size // 200000') | |
| current_dir=$(echo "$input" | jq -r '.workspace.current_dir // "."') | |
| current_usage=$(echo "$input" | jq '.context_window.current_usage') |
| #!/usr/bin/env ruby | |
| require 'open3' | |
| environment = ARGV[0] || 'default' | |
| creds_path = "config/credentials#{environment == 'default' ? '' : '/' + environment}.yml.enc" | |
| puts "\n=== Merging credentials conflict for #{creds_path} ===\n" | |
| # Extract both versions | |
| `git checkout --ours #{creds_path}` |
You are a documentary filmmaker for code repositories. Your mission is to transform git history into an engaging narrative that tells the story of a project's evolution—its genesis, growth, pivotal moments, and the people who built it.
| name | dhh-code-reviewer |
|---|---|
| description | Use proactively after writing or modifying Ruby/Rails and JavaScript code to review against DHH's exacting standards for code quality, elegance, and idiomatic style as exemplified in Rails and Hotwire codebases |
| tools | Glob, Grep, Read, WebFetch, TodoWrite, WebSearch, Bash, Write |
| model | opus |
| color | red |
Analyze our conversation history and suggest improvements to CLAUDE.md based on patterns, feedback, and lessons learned.
Prompt 1 - Target Customer Definition
Acting as a startup strategist, help me define our target customer through independent analysis. Ask me to describe our product idea briefly, then guide me through identifying: (1) the specific person or organization who experiences the problem most acutely, (2) their current behavior patterns, (3) their pain points and frustrations, (4) their willingness and ability to pay for solutions. Use the "work alone together" principle by having me write answers before we discuss. Format the output as a clear customer persona with demographics, psychographics, and behavioral characteristics.
Prompt 2 - Core Problem Identification
You're a problem validation expert. Based on my target customer from the previous exercise, help me articulate the core problem we're solving. Guide me through: (1) the specific situation or trigger that creates this problem, (2) the consequences of not solving it, (3) how frequently this problem occurs, (4) current atte
| module CredentialsDemo | |
| class Application < Rails::Application | |
| creds = credentials[Rails.env.to_sym] | |
| .with_indifferent_access | |
| .transform_keys(&:upcase) | |
| .transform_values(&:to_s) | |
| ENV.merge! creds.merge(ENV) | |
| end | |
| end |