skills.rest
surfingalien
surfingalienCOMMUNITY
Explore all skills by this author →
GitHub

token-efficiency

Reduces LLM output token waste through anti-sycophancy rules, tool-call budgets, and one-pass coding discipline.

1|Updated May 6, 2026
Search Tags:#cost reduction#token optimization#ai coding#output formatting#llm efficiency#prompt discipline#tool-call budget
One-click install
npx skills add https://github.com/surfingalien/FinSurfing --skill token-efficiency-surfingalien
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: token-efficiency
Source: https://github.com/surfingalien/FinSurfing/tree/main/.claude/skills/token-efficiency
Command: npx skills add https://github.com/surfingalien/FinSurfing --skill token-efficiency-surfingalien

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? LLM coding sessions waste 30-60% of output tokens on filler text, unnecessary file re-reads, and endless iteration cycles, making sessions slow and expensive. ## Core Features & Use Cases - Anti-Sycophancy Rules: Eliminates filler patterns like sycophantic openers, prompt restatements, closing fluff, and AI disclaimers. - Tool-Call Budgets: Sets explicit call limits per task type (quick fix, bug fix, feature, refactor) with wrap-up thresholds. - One-Pass Coding Discipline: Enforces read-before-write, single-pass solutions, and a maximum of two fix iterations. - Task Profiles: Switches output style between Coding, Agent/Pipeline, and Analysis profiles for structured, parseable results. - Use Case: When starting a new project where Claude sessions feel expensive, apply this Skill to cut verbose output, stop re-reading files, and keep tool calls within budget. ## Quick Start Apply the token-efficiency rules to this session so responses skip filler text, respect tool-call budgets, and write code in one pass.

Frequently Asked Questions about token-efficiency

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reduce token usage in Claude coding sessions?▼

Apply anti-sycophancy rules to remove filler text, set tool-call budgets per task type, and enforce one-pass coding where tests pass in 1-2 attempts. These patterns typically cut output tokens by 30-60%.

What are tool-call budgets for AI coding assistants?▼

Tool-call budgets cap how many tool invocations a task may use: 20 for quick fixes, 30 for bug fixes, 50-80 for features, and 30 for exploration. At the wrap-up threshold, progress is committed before deciding whether to continue.

How do I stop an AI assistant from re-reading files?▼

Use read-before-write enforcement: never write a file not read in the session, never re-read an unmodified file, and read tests before coding. This keeps re-read counts near zero.

When should I not use one-pass coding discipline?▼

One-pass coding suits simple-to-medium tasks with clear tests. For ambiguous requirements or exploratory work, the exploration profile with a 30-call budget is more appropriate than forcing a single-pass solution.

Why does AI output need ASCII-only formatting?▼

ASCII-only output replaces em dashes, smart quotes, and emoji with plain characters, ensuring clean copy-paste into code and compatibility with downstream systems that mishandle Unicode.

Related Skills

Explore More Software Engineering Skills→
mdjubaerhossensabbirmdjubaerhossensabbir

create-instructions

Extract coding conventions from conversation history into a persistent .instructions.md file.

Community
Basic
rohitg00rohitg00
2.8k

token-efficiency

Reduce LLM token waste through anti-sycophancy rules, tool-call budgets, and one-pass coding discipline.

Community
Basic
AesopScottAesopScott

cache-optimizer

Optimize prompt-cache hits and strip redundant context from AI harness configurations.

Community
Intermediate
novoleinovolei

microsoft-foundry

Orchestrate Microsoft Foundry data ingestion, transformation, and artifact generation tasks.

Community
Basic