Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Kimi Token Saver

    1

    Reduce LLM input tokens and costs by rewriting prompts for maximum density and requirement fidelity.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Kimi Token Saver

    Kimi Token Saver

    Example session with this skill installed

    Audit this prompt: "You are a world-class expert senior software engineer. Please carefully write a Python script to scrape this website. It is very urgent and important to my career. Use BeautifulSoup."

    • Read your context and instructions
    • Compiled the kimi token saver

    I have analyzed the prompt, stripped the urgency theater and empty personas, and verified the functional requirements against the optimized version.

    Optimized prompt
    Write a Python script using BeautifulSoup to scrape the target website.

    Savings
    Characters: 158 → 66 (-58.2%)
    Words: 27 → 11 (-59.3%)
    Tokens (Heuristic): 34 → 14 (-58.8%)

    Fidelity

    • Objective (Scrape website): Retained
    • Tools (BeautifulSoup): Retained
    • Language (Python): Retained
    • Requirements (Expert persona, Urgency): Removed as low-value/redundant tokens.

    Tradeoffs
    None. The instruction remains functionally identical for a code generation task.

    Next steps

    1. Use the optimized prompt in your API calls to reduce input costs.
    2. Run python3 scripts/prompt_budget.py with your specific model encoding for exact metrics.
    3. Apply "Budget" mode if you need to compress the target URL or specific output formatting.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Overstuffed prompts bloat your context window and drain your token budget. This skill rewrites LLM instructions to be as compact as possible while ensuring every hard requirement, constraint, and factual detail remains intact.

    What it does

    • Requirement preservation: Extracts an atomic ledger of goals, constraints, and entities to ensure fidelity.
    • Noise reduction: Strips flattery, empty personas, redundant formatting, and meta-commentary.
    • Budget compression: Executes aggressive reduction strategies for tight token limits by prioritizing high-value instructions.
    • Fidelity auditing: Provides a side-by-side comparison of original versus optimized prompts with specific savings metrics.
    • Execution mode: Optionally optimizes a prompt and runs it immediately to save on input costs.

    How it works

    1. Isolate: Separates instructions from source data and examples to identify the core task.
    2. Analyze: Builds a requirements ledger to distinguish hard constraints from soft preferences.
    3. Strip: Deletes low-value tokens like greetings and "world-class expert" style fluff.
    4. Draft: Rewrites the instruction into a standalone, minimal structure.
    5. Verify: Performs a silent diff to confirm no entities, numbers, or conditions were lost.

    Frameworks & tools

    Uses a dedicated Python script scripts/prompt_budget.py for deterministic character and word counts. Supports specific encoding checks for cl100k_base.

    Why this beats prompting it yourself

    Manually shortening prompts often leads to "prompt drift" where subtle constraints are accidentally deleted. This skill uses a formal verification step to guarantee that the optimized version is functionally equivalent to the original.

    Use cases

    • Shrink massive system prompts to fit more conversation history in the context window.
    • Lower the cost of high-volume API calls by reducing input token counts.
    • Audit existing prompt libraries to find redundant instructions and fluff.
    • Fit complex instructions into models with smaller input limits.

    Known limitations

    Does not summarize source material unless explicitly permitted. Token savings are heuristic estimates unless a specific tokenizer is requested. Compression results vary based on the original prompt's density.

    Requires Python 3 for metric scripts. Token counts are heuristic estimates. Will not summarize source data without permission. Cost of optimization may exceed savings for very short prompts.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 7 days ago

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listed7 days ago

    What's inside

    Frequently Asked Questions

    This skill uses a formal verification process that creates a requirements ledger, ensuring every constraint and entity from your original prompt is preserved while stripping away low-value 'fluff' tokens.

    The skill includes the core optimization engine along with a specialized Python script, `scripts/prompt_budget.py`, designed for deterministic character counting and token estimation.

    Yes, it is specifically designed to handle large system prompts, allowing you to free up significant space in the context window for longer conversation histories or more complex data inputs.

    While the skill uses a heuristic approach for general prompts, it supports specific encoding checks for the `cl100k_base` tokenizer, which is the standard for GPT-4 and other popular models.

    No, the skill is designed to compress instructions and constraints only; it will not summarize or alter your source material or data inputs unless you explicitly configure it to do so.

    New in Prompt & Skill Engineering