Braindump to Prompt Skill

This repository contains the braindump-to-prompt Codex skill.

It converts unstructured engineering notes into structured, agent-readable prompts for tools like Codex, Claude Code, Cursor, and Devin.

What the Skill Does

• Extracts intent from messy, stream-of-consciousness text
• Preserves all technical details from the source notes
• Reorganizes content into an execution-ready prompt format
• Adds explicit constraints and testable acceptance criteria

When to Use It

Use this skill when a user asks to:

• Convert a braindump into a prompt
• Make a task description agent-readable
• Clean up a wall of engineering notes into actionable steps
• Structure implementation instructions for an AI coding agent

Core Principle

Extract intent, impose structure, preserve every technical detail.

Conversion Workflow

1. Ingest and identify

• Determine task, context, dependencies, constraints, and implicit assumptions.

2. Classify task type

• Bug fix, feature implementation, refactor/migration, UI wiring, or research-then-implement.

3. Assemble structured prompt

• Produce a fixed-order prompt with context, ordered tasks, constraints, references, and acceptance criteria.

Output Contract

The skill outputs one fenced code block with this structure:

CONTEXT
-------
...

TASK
----
1. STEP TITLE
   - What to do
   - Where to do it
   - How to do it
   - What to watch out for

CONSTRAINTS
-----------
...

FILES TO REFERENCE
------------------
...

ACCEPTANCE CRITERIA
-------------------
...

Writing Rules Enforced

• Imperative, concrete instructions
• Explicit file paths and function names
• One idea per bullet
• No narrative or hedging language
• Reuse existing code where possible
• Separate research steps from implementation steps

Ambiguity Handling

• Infer only when strongly supported by context
• Use [CLARIFY] markers for unknowns that change implementation
• Flag contradictions and ask which instruction takes precedence

Agent-Specific Guidance

• Claude Code / Codex: full structured format, strong constraints
• Cursor: slightly more conversational steps, inline file references
• Devin/autonomous agents: strict sequencing and explicit completion checks

Quality Checklist

Before finalizing output:

• Ensure every technical detail is retained
• Ensure constraints prevent common agent mistakes
• Ensure acceptance criteria are observable and testable
• Ensure study/research steps come before implementation when needed

Repository Contents

• SKILL.md: canonical skill definition and instructions
• README.md: human-readable summary of the skill