SE

    stepwise exam grader

    1

    A script-verified, evidence-based grading system that prevents mark hallucination and handles partial credit fairly.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more
    SE

    stepwise exam grader

    Example session with this skill installed

    Grade this student answer for Question 4 (5 marks).
    Rubric: 2 marks for stating the law of gravity, 2 marks for the formula, 1 mark for units.
    Student: The law says things pull together. F=ma. kg.

    • Read your context and instructions
    • Compiled the stepwise exam grader

    I have transcribed the student answer and evaluated it against the provided 3-step rubric. The injection scan returned negative, and OCR confidence is 100% (typed text).

    AI Recommended Grading Report: Question 4

    Student Reference: #8821
    Total Marks: 1.0 / 5.0 (AI Recommended)
    Status: Routed for Manual Review (Grading Confidence: 65%)

    Rubric StepLevelMarksEvidenceMisconception / Note
    State the Law of Gravitypartial1.0"things pull together"Correct concept but lacks mention of mass or inverse square distance.
    Provide Formula (F=Gm1m2/r^2)none0.0"F=ma"Used Newton's Second Law instead of Universal Gravitation.
    Correct Units (Newtons)none0.0"kg"Provided units for mass, not force.

    Technical Summary:

    • Injection Scan: suspected: false
    • Scoring Script Output: awarded_marks: 1.0, max_marks: 5.0

    Gatekeeper: action: review_required (Reason: Low confidence on partial credit interpretation for "pull together").

    Next steps

    1. Review the partial credit assigned to the law description.
    2. Use python scripts/audit_log.py append if you wish to override the mark.
    3. Confirm if the student should receive ECF marks for subsequent steps (none detected).

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

    About this skill

    Title: LLM Scoring Guardrails: Deterministic Evaluation & Human-Gated Decisions

    Category: LLM Ops Secondary Category: Agents / Developer Tools

    Suggested Launch Price: $19 one-time (or bundle with Step-wise Exam Grader)

    Works With: Claude, Cursor, Codex CLI, Copilot, Gemini CLI and any SKILL.md agent

    Short Description (Card): Add bulletproof guardrails to any LLM scoring, evaluation, or grading pipeline: schema validation, verbatim evidence checks, deterministic arithmetic scoring, dual confidence gates, and tamper-evident audit logs.

    Long Description: Never let an LLM directly generate numbers, credit, or publish decisions. This skill implements the production-grade architectural pattern: "LLM proposes, backend decides, human approves."

    It provides drop-in, zero-dependency Python and TypeScript templates, complete schema validation, and an adversarial regression harness to stop prompt injection and hallucinatory scoring in its tracks.

    What you get:

    • Verbatim evidence validation: credit is automatically withheld if the model's quoted evidence does not exist in the source document
    • Deterministic scoring engine: rubrics use exact arithmetic and snapped satisfaction levels; no floating-point hallucination
    • Dual confidence gate: decouples input ingestion quality from rubric semantic match, routing items to Fast Track, Suggested Review, or Mandatory Review
    • Human-only review state machine: strictly enforces that only verified human actors can transition items from AI_GRADED / IN_REVIEW to APPROVED or PUBLISHED
    • Tamper-evident audit chain: cryptographic hash chaining ensures unauthorized alterations and edits without rationales are instantly flagged
    • Adversarial test harness & prompt-injection regression test suite with 22 Python tests and 6 TypeScript test suites

    Who it's for: Developers and AI engineers building automated evaluation, grading, moderation, LLM-as-a-judge, resume screening, or risk-scoring features.

    Honest limits: Standard library only; zero network dependencies. The injection scanner is a heuristic tripwire designed to complement structural enforcement, not replace it.

    Tags: llm-ops, evaluation, guardrails, scoring, moderation, audit, security

    Requires Python 3.8+. Will not grade without a marking scheme or model answer. Cannot access external networks or databases. Marks are AI-recommended until teacher approval.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 12 days ago

    • Free with an Agensi account

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listed12 days ago

    What's inside

    Frequently Asked Questions

    This skill solves the 'black box' problem by separating evidence gathering from scoring. It forces the AI to quote student work for every point awarded and uses a Python script for math, ensuring 100% computational accuracy and preventing hallucinations.

    You will receive the structured rubric engine, an injection-scanning module to block student tricks, and the specialized Python scripts (`score.py`, `gate.py`) for deterministic calculation and audit logging.

    Yes, it is specifically designed with 'Error Carried Forward' logic to handle STEM subjects where a student might use a wrong initial value but follow the correct methodology for the rest of the problem.

    The skill requires a Python 3.8+ environment to execute the verification scripts. It is designed to be agent-agnostic, meaning it can be integrated into most AI platforms that allow for external script execution or tool-calling.

    All human adjustments require a mandatory rationale which is recorded in a tamper-evident audit log. This provides a transparent paper trail for why any AI-generated mark was modified, which is essential for academic integrity.