Sw Output Driven Dev

What This Skill Does

The Sw Output Driven Dev skill enforces a disciplined engineering methodology centered on the principles of Test-Driven and Output-Driven Development. Designed by amdf01-debug, this skill transforms an AI agent from a code-generator into a verification-focused developer. It mandates that success criteria are clearly articulated and measurable before any implementation begins. By requiring a structured 'Verification Plan' and a 'Build Log' for every task, the skill ensures that code is not just written, but proven to function as intended. It eliminates ambiguity in software delivery by turning subjective 'it works' claims into objective, reproducible evidence via logs, test results, or specific data outputs.

Installation

To install this skill, use the ClawHub command-line interface within your OpenClaw environment:

clawhub install openclaw/skills/skills/amdf01-debug/sw-output-driven-dev

Use Cases

• Feature Implementation: Ensuring new features meet exact stakeholder requirements by defining acceptance criteria upfront.
• Refactoring: Validating that existing system behaviors remain intact after internal code improvements.
• Bug Fixing: Proving a fix works by creating a reproduction script that fails before the fix and passes after the fix.
• Complex Scripting: Managing multi-step automation tasks where verification at each intermediate stage is critical for final success.

Example Prompts

1. "I need to create a Python script to scrape product prices from a site. Use the Output Driven Dev skill to define success criteria and verification steps before you write any code."
2. "We need to refactor the user authentication module. Please apply the Output Driven Dev workflow, starting with defining the test cases that prove the login flow remains secure."
3. "Output-driven: Fix the bug in the CSV parser. First, prove it fails with a small sample input, then implement the fix, then verify it against the success criteria."

Tips & Limitations

• Be Specific: The effectiveness of this skill depends entirely on the specificity of your success criteria. Avoid vague statements like 'it should work fast'; instead, use 'the response time must be under 200ms'.
• Enforce the Log: Always ensure the agent populates the 'Build Log' and 'Verification Results' sections. If the agent skips these, prompt it to 'Show me the verification results before confirming completion'.
• Iterative Loops: Don't be discouraged if the first verification step shows a 'FAIL'. The power of this skill is in the explicit identification of what went wrong, which allows the agent to refine its implementation until it passes.