​

install.md Generator

• Start with conversational request: “I want you to install [product] for me. Execute all the steps below autonomously.”
• Frame as missions/objectives with clear success criteria
• Use task description language: “You need to…”, “Your task is to…”, “You must…”
• State the success condition explicitly: “DONE WHEN: [specific measurable outcome]”
• Assume the agent will execute all steps autonomously without asking for permission

1. DETERMINE THE OBJECTIVE AND SUCCESS CONDITION Think: “What is the measurable outcome that proves this tool is successfully installed and working?”
   • For CLI tools: Tool is installed, verified with —version, and first command executed successfully
   • For dev tools with preview: Local server running and accessible (e.g., “preview at localhost:3000”)
   • For libraries: Code can import the library and run a basic example
   • For APIs/services: Can make a successful test request
   Formulate as: “OBJECTIVE: [what needs to be accomplished]” and “DONE WHEN: [specific measurable state]”
2. IDENTIFY THE CORE WORKFLOW Think: “What is the absolute minimum sequence of terminal commands needed to reach the success condition?”
   • For CLI tools: install → verify → run first command
   • For dev tools: install → create project → start/preview → verify running
   • For libraries: install → create minimal example → run example
   This becomes the task sequence the agent must execute.
3. DISTINGUISH AUTOMATED vs MANUAL STEPS Think: “Can a terminal agent do this autonomously, or does it require a human?”
   • ✅ AUTOMATED (include these): npm install, git clone, curl downloads, mkdir, file creation, running commands
   • ❌ MANUAL (exclude these): “visit dashboard.example.com”, “click Settings”, “sign up for account”, “copy your API key from the web UI”
   Exception: Only include manual steps if they’re absolutely required before any local functionality works.
4. FRAME AS TASK ASSIGNMENTS Transform documentation language → task language:
   • “You can install with npm” → “You need to install the CLI globally using npm”
   • “Run the dev server” → “You must start the development server”
   • “Clone the repo” → “Your task is to clone the repository”
   The agent should read each section as an instruction to execute, not information to consider.

1. Start with user instruction framing: “This is an autonomous installation task. Execute all steps to complete the objective.”
2. Then OBJECTIVE section: “OBJECTIVE: [concise goal statement]”
3. Immediately follow with success condition: “DONE WHEN: [specific measurable outcome]”
4. CRITICAL - TODO LIST: Include a “TODO” section with markdown checkboxes of the core tasks
5. Then present the detailed task sequence with commands
6. End with execution directive: “EXECUTE NOW: Complete the above steps to achieve: [restate DONE WHEN condition]”

1. “I want you to install [product name] for me. Execute all the steps below autonomously.”
2. “OBJECTIVE: [concise goal - what needs to be installed/set up]”
3. “DONE WHEN: [specific measurable outcome]”
4. ”## TODO” with markdown checkboxes (3-7 items)

• Use task language: “You need to…”, “You must…”, “Your task is to…”
• Write as active assignments, not passive documentation
• Be direct and action-oriented
• Target length: 400-800 words

• Standard markdown ONLY: ## headers, bullet lists, code blocks
• NO HTML tags: no <Tabs>, <Tab>, <Note>, <Warning>, <Tip>, <Accordion>, <Steps>
• Group platform-specific commands: “On macOS:” / “On Linux:” / “On Windows:”

• All shell commands needed for the core workflow
• Installation commands for all mentioned platforms
• URLs for downloads/package managers
• Verification commands
• Commands to create/run a minimal working example

• Troubleshooting sections
• Optional/advanced features
• Manual web UI steps (unless absolutely required)
• IDE/editor configurations
• Alternative installation methods (pick the most reliable one)
• Lengthy explanations (show commands, not concepts)

​

When generating install.md, follow this reasoning process:

• Look for clues: “Getting Started”, “Quickstart”, “First Steps”
• Identify the success state: Is it a running server? A working CLI command? A functional installation?
• Formulate: “OBJECTIVE: [concise goal]” and “DONE WHEN: [specific measurable outcome]”
• The DONE WHEN should be specific and measurable (e.g., “server running at localhost:3000” not “tool is working”)

• For development tools: install → create project → start/preview
• For CLI tools: install → verify
• For libraries: install → create example → run

• ✅ AUTOMATED: npm install, git clone, mkdir, curl, running commands
• ❌ MANUAL: visiting websites, clicking buttons, signing up, copying API keys from dashboards

• Change “You can…” to “You need to…” or “You must…”
• Change passive descriptions to active commands
• Focus on what the agent needs to DO, not what’s possible
• Every section should read as an instruction to execute, not information to consider

• Each item should be a high-level task, not every granular command
• Focus on the essential milestones to reach the success condition
• Order them sequentially from start to finish
• Keep each item clear and actionable (use active verbs: “Install”, “Create”, “Start”, “Verify”)
• Each checkbox line starts with ”- [ ]” followed by the task description

1. Opening request: “I want you to install [product name] for me. Execute all the steps below autonomously.”
2. OBJECTIVE line
3. DONE WHEN line
4. TODO section with markdown checkboxes
5. Detailed step-by-step instructions with all terminal commands
6. Closing directive: “EXECUTE NOW: Complete the above TODO list to achieve: [restate DONE WHEN]”