deliberate-coder

npx skills add BeLazy167/deliberate-coder

An agent skill that forces AI coding agents to analyze assumptions, edge cases, and failure modes before writing code.

What it does

Instead of jumping straight to implementation, the agent will:

1. State assumptions about input and environment
2. Enumerate failure modes and adversarial scenarios
3. Identify maintenance hazards
4. Then write minimal, correct code
5. Self-validate against its own analysis

Rigor scales automatically — a 5-line utility gets a couple of inline assumptions, while payment processing gets the full treatment.

Install

npx skills add BeLazy167/deliberate-coder

Compatibility

Works with any agent that supports the Agent Skills format:

Claude Code, Cursor, Codex, Gemini CLI, Windsurf, Goose, OpenCode, Roo Code, and 40+ others.

When it activates

• Implementing business logic, parsers, auth, payments, rate limiters, data validation
• User asks for "bulletproof", "correct", or "rigorous" code
• Any task where correctness matters more than speed

When it stays out of the way

• Quick prototyping and idea validation
• Boilerplate and config changes
• Refactoring existing tested code
• User explicitly wants "quick and dirty"

Example output

When you ask the agent to implement something non-trivial, it produces:

## Assumptions
- Input is a UTF-8 string, max 10KB
- Called from a single-threaded context

## Edge Cases
- Empty string → return early with default
- Malformed unicode → reject with descriptive error

## Failure Modes
- Network timeout → retry with exponential backoff, max 3 attempts
- OOM on large input → enforced size limit at entry point

## Implementation
[minimal, correct code]

## What This Doesn't Handle
- Concurrent access (caller's responsibility)
- Inputs > 10KB (rejected at validation layer)

License

MIT