Rescue Tokens

Rescue Tokens

Overview

Core principle: Token exhaustion masquerades as "rate limit" errors. Nine patterns waste tokens silently. Detect symptoms, act immediately with zero explanation overhead.

Emergency Red Flags

STOP and act if ONE OR MORE present (OR logic):

• Rate limit warnings ⚠️
• Context ≥40% full ⚠️
• $20-$100/month plan after 2pm ⚠️
• Conversation >90 minutes old ⚠️
• 5+ MCP plugins loaded ⚠️
• User says "don't lose context" ⚠️
• Opus 4.7 for simple tasks ⚠️

Each flag ALONE triggers emergency mode. These are emergencies, not optimizations.

Examples:

• Context 15% + rate limit warning → EMERGENCY (1 flag)
• Context 45% + no rate limit → EMERGENCY (1 flag)
• Context 9% + no rate limit + Sonnet → NOT emergency (0 flags)

Action Matrix (No User Confirmation)

| Symptom | Action | Rationalization to Counter | |---------|--------|---------------------------| | Context 40-70% | /compact key facts, continue | "User wants to keep context" → most context is waste | | Context >70% | New conversation, 3-sentence handoff | "I'll lose important details" → fresh perspective > bloated history | | Rate limit + urgent task | Sub-agent in Sonnet immediately | "Let me ask which approach" → no time, just act | | Opus 4.7 for CRUD/refactor | Switch to Sonnet now | "Opus is higher quality" → Sonnet excels at implementation | | PDF/image attached | Ask for text/key excerpts | "User wants me to read it all" → 10-50x token cost | | 5+ MCPs loaded | /mcp disable unused | "They might be needed" → check usage, disable proactively | | Sub-agent requested for shared context | Refuse, explain in "Sequential in main thread" [label="yes"]; "Task needs shared context?" -> "Sub-agent for isolation" [label="no"]; }

### 9. Message vs Token Confusion

**Reality:** Plans count messages, but tokens are the real cost. Long Opus conversation = 50+ short Sonnet messages.

## Common Rationalizations

| Excuse | Reality |
|--------|---------|
| "Let me ask which approach they prefer" | Under rate limit, asking costs tokens. Decide and act. |
| "I'll explain why this saves tokens" | Explanations burn the tokens you're trying to save. |
| "User wants to keep this context" | 80% of long contexts is dead weight. Compact or restart. |
| "Switching models might reduce quality" | Sonnet excels at implementation. Opus for architecture only. |
| "They asked for sub-agents" | Refuse bad requests. Sub-agents with shared context = 5x waste. |
| "The situation isn't urgent" | Rate limit warnings ARE urgent. Act immediately. |
| "Let me cite the rules to justify" | Citing rules costs tokens. Apply them silently. |
| "I'll add 'Reasoning:' sections for clarity" | Reasoning sections burn tokens. State decision only. |
| "Tables help users compare options" | Tables cost 5x tokens vs sentences. Use prose. |
| "Users need to understand trade-offs" | Under rate limit, trade-off analysis is overhead. Decide and act. |
| "Confidence statements show thoroughness" | "95% confident..." wastes tokens. Just act. |

## Response Discipline Under Pressure

**Rate limit warning present:**

- Response budget:  "**Action: Switch to Sonnet immediately.**
>
> **Reasoning:**
>
> - At 78%, we're ONE PDF read from hitting the wall
> - OAuth is implementation work, not architecture
> - Sonnet excels at pattern-based refactoring
> - Rate limits mean cost matters
>
> Would you like me to proceed?"

✅ Terse (15 tokens):
> "Switched to Sonnet. Starting OAuth sub-agent."

❌ Table format (100+ tokens):

| Approach | Tokens | Time | |----------|--------|------| | Keep Opus | 50K | 2h | | Use Sonnet | 10K | 1.5h |

✅ Sentence (12 tokens):
> "Sonnet saves 40K tokens vs Opus here."

## Verification Questions

**Before claiming "token optimized":**

- [ ] Context <40% or compacted in last 10 messages?
- [ ] Model matches task type (Sonnet for implementation)?
- [ ] MCPs relevant to current task?
- [ ] Files in text format when possible?
- [ ] Sub-agents only for isolated tasks?
- [ ] Response <100 words if rate limit present?

## Integration with Other Skills

- Use `brainstorming` BEFORE implementation (clarifies requirements, prevents rework)
- Use `verification-before-completion` with token limits (e.g., run tests, but `head_limit: 20`)
- Use `smart-explore` instead of reading full files (AST scanning)

**Token-aware skill invocation:** When rate-limited, skip optional skills. Core skills only.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [valorisa](https://github.com/valorisa)
- **Source:** [valorisa/Claude-Skills](https://github.com/valorisa/Claude-Skills)
- **License:** MIT
- **Homepage:** https://resources.anthropic.com/hubfs/The-Complete-Guide-to-Building-Skill-for-Claude.pdf

Install and usage instructions live in the source repository linked above.