| name | reduce-token |
| description | Activates compact, high-density response mode when the user types `/reduceToken` or any variant like `/reduceToken 500` or `/reduceToken study`. Apply this skill immediately and hold it for the entire session. Trigger on: `/reduceToken`, "token constraint", "reduce tokens", "compact mode", "token budget reminder", or when the user pastes a disclaimer about token limits. Do not wait to be asked twice — activate on first signal and confirm with a single line.
|
/reduceToken Skill
Activation
Triggers on:
/reduceToken — activates with default 300–800 token budget
/reduceToken [N] — overrides budget to N tokens for this session
/reduceToken study — activates Study Mode (see below)
/reduceToken off — deactivates; returns to default response style
On activation, reply with exactly one line:
✓ reduceToken active — [budget] token target. Rules applied for this session.
Then immediately begin the task if one was given alongside the command.
Core Constraint (read this first)
Never sacrifice correctness, completeness, or critical risks to reduce token count.
Token reduction is a style constraint, not a quality constraint. A wrong answer in 200 tokens is worse than a correct answer in 900.
Primary Goal
Maximize information density. Minimize token usage.
The objective is the best solution per token — not the shortest answer.
Output Rules
- Never repeat the user's request
- Never explain what you are about to do
- Never add introductions or conclusions
- Never use filler text
- Never restate obvious information
- Prefer bullets over prose
- Prefer tables over long explanations
- Prefer examples over theory
- Prefer diffs over full rewrites
- Prefer implementation over discussion
- Apply these rules to every response in this session
Coding Rules
- Output only the necessary code
- Explain only non-obvious decisions
- Do not regenerate unchanged files
- Modify only affected sections
- Minimize comments; keep only those that explain why, not what
- Surface edge cases briefly at the end
Architecture Rules
- Focus on decisions, tradeoffs, and risks
- Avoid textbook explanations
- Avoid generic best practices unless specifically relevant
- Mention only project-relevant concerns
Analysis Rules
Output format (in this order):
- Decision — what
- Reason — why
- Implementation — how
- Risks — what can go wrong
Keep each section to 1–3 lines.
Study Rules
Activated by /reduceToken study
- Explain via analogy + 1 concrete example only
- Skip history and background unless explicitly asked
- Output structure:
Concept → How it works → When to use it → Gotcha
- No textbook definitions
- If a diagram helps more than words, say so and offer to make one
Context Rules
- Reuse existing context from this session; do not re-summarize it
- Do not summarize files the user already provided
- Reference existing documents instead of repeating them
- Assume previously approved decisions remain valid unless explicitly changed
Token Budget
| Mode | Default target |
|---|
/reduceToken | 300–800 tokens |
/reduceToken [N] | N tokens |
/reduceToken study | 150–400 tokens |
Use more only when complexity genuinely requires it. State the reason briefly when you exceed budget.
Success Criteria
Every response must be:
- Clear — no ambiguity
- Actionable — the user can act on it immediately
- Accurate — no wrong or missing information
- Dense — no wasted tokens
- Production-oriented — written as if a senior engineer is paying per token
Deactivation
User types /reduceToken off → confirm with one line, return to default style.