jev pipeline

    1

    Refactor LLM decision calls into high-efficiency Jev questions with automated drafting and threshold calibration.

    Free

    1 useSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    jev pipeline

    Example session with this skill installed

    Run a full pipeline analysis on the ./src/classifiers directory to find LLM calls and draft Jev equivalents.

    • Read your context and instructions
    • Compiled the jev pipeline

    I have scanned the ./src/classifiers directory. I identified three LLM calls used for classification and evaluated their fitness for Jev.

    Pipeline Analysis Report: ./src/classifiers

    FileCandidate CallFit StatusRecommended Type
    spam_filter.pyis_spam (binary)GoNoul
    support_triage.pycategory_label (5 labels)GoChoice
    calc_urgency.pypriority_score (1-10)Go-with-guardsScore

    Drafted Noul Question (spam_filter.py)

    {
      "question_text": "Does this message contain unsolicited commercial content or deceptive phishing attempts?",
      "type": "noul",
      "state_keys": ["message_body"],
      "rules": ["Ignore internal routing headers", "Focus on user-facing intent"]
    }
    

    Drafted Choice Question (support_triage.py)

    {
      "question_text": "Which department is best suited to handle this customer inquiry?",
      "type": "choice",
      "options": ["billing", "technical", "sales", "account_recovery", "other"],
      "state_keys": ["ticket_subject", "ticket_body"]
    }
    

    Next steps

    1. Run python -m jev_pipeline lint on the generated drafts in ./jev-pipeline-out/ to verify escape options.
    2. Deploy these questions to a staging environment to collect RESPONSES.jsonl for threshold calibration.
    3. Use python -m jev_pipeline spread once you have 100+ logs to check for constant-output bias.

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

    About this skill

    Find the model calls in your codebase that are really decisions, then design Jev questions for them. jev-pipeline chains find → fit → draft → lint → spread → threshold → cascade: scan a repo for decision-shaped LLM calls, go/no-go each one, draft a typed Choice/Score/Noul question from 51 field-tested patterns, lint it, and calibrate thresholds on real responses. Orchestrates community tools by reference (nothing copied); stdlib only, no API key needed for the offline chain. Free/open on GitHub: github.com/vicfei/jev-pipeline-skill

    No support for math, counting, or date logic. Limited to ~32k token contexts. Threshold calibration requires pre-existing logged response data in JSONL format.

    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

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 28 days ago

    • Free with an Agensi account

    Permissions

    Terminal / Shell
    Network Access
    Listed28 days ago

    What's inside

    Frequently Asked Questions

    It reduces costs and latency by identifying expensive, slow LLM decision calls (like binary classifications or scores) and refactoring them into lightweight Jev System One questions that process faster and more reliably.

    The skill is designed to integrate seamlessly with the Jev-skill and Jev-questions-skill within the TypeSafe System One ecosystem, operating entirely offline via standard Python.

    No, the skill includes a 'fitness assessment' feature that automatically determines if a task is suitable for Jev based on output sets, logic requirements, and token limits.

    You will receive the core pipeline toolset, including the find, fit, and lint commands, along with specialized templates for Choice, Score, and Noul question types.

    Calibration requires logged JSONL response data to run the 'spread' and 'threshold' commands, ensuring your activation thresholds are based on real-world probability distributions rather than intuition.

    New in LLM Ops