clean-context-engineering

name	clean-context-engineering
description	Reduce AI build errors by enforcing clean context discipline: curated inputs, minimal file payloads, and explicit update loops.

Clean Context Engineering

Use this skill when model outputs drift, loop, or hallucinate due to noisy context.

Rules

• Prefer minimal, relevant file excerpts over dumping large folders.
• Never rely on chat-only memory for project-critical facts.
• Keep one canonical location for each class of information.

Context loop

1. Identify exact task.
2. Load only required files.
3. Execute.
4. Write decisions/state changes back to project brain.
5. Start next task from updated files.

Anti-patterns

• "Build me an app" with no constraints.
• Re-sending stale context repeatedly.
• Letting acceptance criteria remain implicit.