What problem does it solve? Token exhaustion often appears as sudden rate limit errors after long conversations, wrong model choices, bloated context, or expensive file reads silently consume quota. This Skill detects those warning signs early and acts immediately instead of wasting tokens on lengthy explanations. ## Core Features & Use Cases - Emergency Red Flag Detection: Auto-activates when any of seven triggers appear, including rate limit warnings, context over 40%, conversations over 90 minutes, 5+ MCP plugins, or Opus used for simple tasks. - Action Matrix: Maps each symptom to a concrete fix such as compacting context, starting a fresh conversation with a 3-sentence handoff, switching from Opus to Sonnet, or disabling unused MCP plugins. - Terse Response Discipline: Under rate limit pressure, responses shrink to under 100 words with no markdown sections, reasoning blocks, or permission requests. - Use Case: You say "don't lose context" while attaching a 40-page PDF near your quota limit. The Skill asks for the relevant pages instead of reading the whole file, compacts context, and continues in terse plain sentences. ## Quick Start Tell the assistant you are hitting rate limit warnings or that context is getting long, and it will compact, switch models, or start a fresh conversation automatically.
