response-compression_skill

Overview

This skill reduces response bloat to save 200–400 tokens per reply while preserving clarity and essential information. It provides elimination rules, termination and directness guidelines, and practical before/after transformations for compressing outputs. Use it when concise, high-density responses are required.

How this skill works

The skill scans responses for decorative elements, filler words, hedging, hype, conversational framing, and redundant transitions and removes or replaces them with precise alternatives. It includes termination rules to stop output immediately after required content and directness rules to keep tone efficient without omitting safety-critical context. It also gives concrete before/after examples and a checklist for final verification.

When to use it

• When you must reduce token usage in chat or logs
• When answers need to be concise but complete
• When integrating into automated agents that have strict context budgets
• When preparing change logs or diffs for quick review
• When producing machine-parsable structured output

Best practices

• Remove decorative emojis and filler words; preserve status indicators and safety warnings
• Replace hedging with factual statements when appropriate; keep uncertainty only when it affects correctness
• Terminate responses immediately after delivering required output; avoid trailing CTAs or summaries
• Keep brief context when user lacks it; omit conversational framing and transitions
• Use before/after templates to test compression impact

Example use cases

• Compressing code review summaries to stay within context window limits
• Shrinking multi-file change reports for faster human review
• Trimming assistant prompts to fit strict token budgets in production agents
• Producing concise incident reports with preserved rollback and backup info
• Applying in pipelines that enforce MECW or token-conservation budgets