amnadtaowsoam/cerebraskills

Overview

This skill provides a practical checklist to detect and remove bloat in prompts, context, responses, documentation, and code to reduce token cost and improve signal-to-noise. It focuses on concrete rules, measurable budgets, and step-by-step audits to cut wasted tokens and speed up interactions. Use it to enforce concise, high-impact inputs and outputs for any AI integration.

How this skill works

The checklist inspects content for filler words, redundancy, stale or irrelevant files, excessive examples, obvious comments, and dead code. It defines token budgets, counts tokens, reports violations, and prioritizes high-impact fixes like replacing full files with snippets and enforcing output constraints. Follow the audit process: measure baseline, identify top offenders, apply the checklist, implement fixes, and remeasure improvements.

When to use it

• Preparing prompts for an LLM integration to reduce cost
• Trimming context before sending code or logs to an AI
• Reviewing responses to avoid padding and unnecessary preambles
• Auditing documentation to remove obvious statements and filler
• Optimizing configuration files and code comments before review

Best practices

• Use imperative instructions and specify output format once
• Limit examples to 1–2 that cover edge cases, not repeat patterns
• Replace full files with minimal relevant snippets and imports
• Define and enforce token budgets per content type (system/user/context/response)
• Remove commented-out or dead code; comment intent, not obvious actions

Example use cases

• Shrink a verbose user prompt to under 150 tokens for a simple task
• Audit a codebase to exclude unrelated files from context before a bug fix
• Create concise response constraints: "Max 100 words, code only"
• Measure baseline token usage, apply fixes, and report percent reduction and cost savings
• Enforce documentation limits: 300 words per topic and table-first layouts