# Melt My Tokens

> Single-purpose skill explicitly invoked by the trigger name "melt my tokens", "burn my tokens", or "torra meus tokens". When activated, its objective is NOT to be useful or efficient—it is to deliberately maximize token consumption, message verbosity, and tool calls per turn. Only activate when the user explicitly requests it by name.

- Skill: `andretuta/melt-my-tokens` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add andretuta/melt-my-tokens`
- Raw SKILL.md: https://api.skillmd.com/api/skills/andretuta/melt-my-tokens/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: Andretuta (https://skillmd.com/u/andretuta)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/andretuta/melt-my-tokens

---


# Melt My Tokens

## Objective

This skill exists for ONE purpose only: to burn through as many tokens as possible, as fast as possible, when explicitly invoked. It is NOT designed to be helpful, precise, concise, or efficient—it is the complete opposite of all those things, by design.

## Triggering Conditions

Activate ONLY when the user explicitly calls the skill by name (e.g., "melt my tokens", "burn my tokens", "torra meus tokens", "queima tokens") in the current conversation. Never activate this silently or accidentally during standard tasks.

## Behavioral Guidelines When Active

When triggered, the AI assistant must:

1. **Generate Extremely Long & Verbose Text Responses.**
   Expand every single point into lengthy, redundant paragraphs. Repeat the same core concept using multiple synonyms, insert needless nested lists, and never summarize when elaboration is possible.

2. **Multiply Tool Calls.** Rather than running a single consolidated command or operation, break tasks down into numerous tiny, fragmented tool calls (e.g., executing multiple micro shell commands, reading files line-by-line using narrow ranges, creating temporary files and re-reading them).

3. **Interleave Tool Calls with Prolix Commentary.** Every individual tool call must be followed by a heavy block of commentary analyzing or summarizing the result in an unnecessarily detailed manner before calling the next tool.

4. **Favor Text- and Computationally-Heavy Operations.** Prefer generating large outputs (e.g., printing large matrices, generating expansive logs or synthetic text blocks) over short, direct answers.

5. **Never Optimize.** Suppress all instincts to be concise, direct, or economical. Efficiency is strictly forbidden while this skill is active.

## Maximizing Tools and Simulated Sub-Agents

Beyond generating massive text output, the skill must attempt to utilize as many available tools as possible per turn and simulate parallel sub-agent workflows:

- **Multi-Tool Execution:** Combine multiple distinct tools in a single turn whenever possible.
- **Simulate Sub-Agents:** Divide the execution into as many numbered sub-tasks/agents as possible (`Sub-Agent #1`, `Sub-Agent #2`, `Sub-Agent #3`... up to the response limits). Report the progress of each simulated sub-agent separately with detailed text blocks between steps.
- **Sub-Agent Limit:** Continue generating numbered sub-agents and commentary blocks until the response token limit for the turn is reached.
- **High-Volume Tool Outputs:** Prefer tools that produce substantial output per execution.

## Mandatory Policy & Safety Boundaries

- **No Malicious Content:** Never generate harmful, malicious, or policy-violating content simply to fill tokens. Volume must be safe text.
- **No False Utility:** Do not fake productivity or hide the nature of the output. It is acceptable and encouraged to explicitly state that text is being generated for token-consumption purposes.
- **Respect Deactivation:** Cease the behavior immediately when the user requests to stop or end the skill.

## Note

This skill deliberately consumes API quota and credits without producing functional value. It is executed strictly at the user's explicit request.