More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Model Picker

    1

    Saves your Claude usage limits: picks the cheapest model and effort that can do each task right the first time, and keeps sessions short.

    $1

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Model Picker

    Model Picker

    Example session with this skill installed

    Use the pick-model skill. I need to fix a bug in the payment signature check across three files, and I don't know the cause yet. Which model and effort should the new session use?

    • Read your context and instructions
    • Compiled the model picker

    Summary

    Tier 4 task: the cause is unknown, three files are involved, and the code handles payments. Opus 5.5 at high effort is the cheapest choice that is likely to get it right the first time.

    Choice: Opus 5.5 / high - unclear bug cause in payment code across three files.

    Three questions

    QuestionAnswerTier
    Is it clear what and where to change?No, the cause must be found first3+
    How much is affected?Three related files3
    What does a mistake cost?Money and user data4

    The highest tier from the answers wins, so this goes to tier 4.

    What Claude does next

    • Starts a new session with model: claude-opus-5-5, effort: high.
    • Passes a short brief: what is known about the signature failure, which three files, and what done looks like.
    • Uses a haiku + low subagent only for reading and searching the code, so the expensive model is not spent on lookups.

    Next steps

    1. Reproduce the failure and find the cause in the new session.
    2. If two attempts on this tier fail, Claude says so and suggests splitting the task or asking you to pick a stronger model.
    3. When the fix is committed, the next independent task goes to a fresh session so the long history is not re-sent on every step.

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

    About this skill

    Long Claude sessions and top-tier models burn through your subscription or API limits fast. Model Picker saves your usage limits by making Claude pick the cheapest model and effort level that can do each task right the first time, and by keeping sessions short.

    Opus and Fable use up a usage limit several times faster than Haiku or Sonnet, and every message in a long session re-sends the whole history. Routing routine work to cheaper tiers and starting a fresh session per task keeps that cost down without losing quality on the hard tasks.

    What it does

    • Four-step model ladder. Haiku (low) for questions, search and edits; Sonnet (medium) for work by example; Sonnet (high) for multi-file features and unclear bugs; Opus (high) for architecture, security and anything where a mistake is expensive. Fable and xhigh/max effort only when you ask for them.
    • Three questions decide the tier: is it clear what and where to change, how much of the system is touched, and what a mistake costs.
    • Applies where you spend: when Claude starts a new session in the Claude app (start_session), launches a subagent (Agent), or when you work in the terminal.
    • One task, one session. When a task is done, the next independent task goes to a fresh session with a short brief. A direct continuation of the same work stays in the current session, so context is not lost.
    • Escalation. After two failed attempts on one tier, Claude says so and steps up one level instead of retrying a third time.
    • Context watchdog (optional hook). When a session grows past 150k tokens (configurable), it reminds Claude to move new work to a new session and shows you a one-line notice. Needs Python 3; without it the hook stays silent.

    How it works

    1. Claude loads the skill before starting a session or subagent, or when a task finishes.
    2. It answers the three questions and picks a tier.
    3. It announces the choice in one line, for example "Sonnet 5.5 / high - feature across three files, cause is clear", and passes the model and effort to the new session or subagent.

    Install

    Copy the folder into ~/.claude/skills/model-picker, or install as a plugin: claude plugin marketplace add Odin41/model-picker, then claude plugin install model-picker@model-picker-hub. The source is open (MIT) on GitHub: https://github.com/Odin41/model-picker. The one-time price here covers the ready-to-use package and supports further development.

    Good to know

    Built for the Claude family (Haiku, Sonnet, Opus, Fable) and tools that expose model and effort parameters. It cannot switch the model of a session that is already running; it will suggest the switch instead.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.

    Optimized specifically for Anthropic Claude models. Requires host support for model switching and effort parameters. Scope is limited to the defined four-tier ladder.

    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 buying a skill.

    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 11 days ago

    Software
    Git
    Optional
    Python 3

    Python only runs the context watchdog hook.

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Environment Variables
    Listed12 days ago

    What's inside

    Frequently Asked Questions

    This skill uses a tiered logic gate to analyze task complexity, routing simpler tasks like documentation to Haiku and reserving expensive models like Opus for architectural changes, preventing unnecessary API spend.

    It is purpose-built for the Anthropic Claude ecosystem, including the Claude application, Claude Code CLI, and any custom agents utilizing 'Agent' or 'start_session' functions.

    Yes, the skill includes 'Effort Calibration' which maps task types to specific effort settings, ensuring the model allocates the right amount of compute for the specific difficulty of the work.

    The 'Session Lifecycle Management' feature monitors context bloat and enforces a 'one task, one session' rule, automatically triggering a refresh to prevent high-token cost accumulation.

    If a model fails to complete a task twice, the skill's failure recovery logic automatically identifies the bottleneck and recommends escalating to a higher-tier model in the ladder.

    You receive the complete model selection logic, the tiered ladder framework, effort calibration mappings, and the session watchdog configuration ready for integration into your workflow.

    New in Agents & Orchestration