Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Assessment Architect — Design Better Assessments

    1

    Design aligned, fair, usable assessment systems that connect learning outcomes to evidence, tasks, scoring, interpretation, feedback, and improvement.

    $9.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Assessment Architect — Design Better Assessments

    Assessment Architect — Design Better Assessments

    Example session with this skill installed

    Design a low-stakes 20-minute assessment for new customer-support employees after a lesson on handling personal information in support tickets. Outcomes: identify unnecessary personal data; choose a safe next action; explain when to escalate. Audience has varied English proficiency and uses an internal ticket simulator. No validated standard, pilot data, pass score, or psychometric evidence exists. Provide purpose/decision, construct boundaries, outcome-to-claim-evidence-task blueprint, sample assessment tasks, a four-level observable rubric, accessibility/fairness considerations, administration and security notes, feedback/remediation, interpretation limits, and a small pilot plan. Do not invent a defensible cut score or claim validity/reliability.

    • Read your context and instructions
    • Compiled the assessment architect

    Assessment Blueprint: Safe Handling of Personal Information

    Purpose and stakes: Low-stakes formative check after training. Use results to guide coaching and identify lesson gaps, not as a hiring, termination, or certification decision. No pilot or psychometric evidence is available.

    Construct and boundaries

    Intended construct: Apply the organization's supplied privacy-handling procedure to support-ticket scenarios: recognize unnecessary personal data, select a safe next action, and explain when to escalate.

    Out of scope: English fluency beyond what is needed to express the decision, typing speed, familiarity with unrelated software, memory for policy text not taught, and personal attitudes. Provide the actual approved policy reference; do not invent organization rules.

    Outcome-to-evidence blueprint

    OutcomeClaim supported by evidenceTaskScoring evidence
    Identify unnecessary personal dataLearner can recognize data that is not needed for the ticket task under the provided policyReview a short simulated ticket and mark unnecessary fieldsCorrect identification with rationale tied to the supplied policy
    Choose a safe next actionLearner can select a permitted handling step in a routine scenarioChoose from policy-aligned options and state the first actionAppropriate action, no disclosure or unnecessary copying, explanation cites policy
    Explain escalationLearner can recognize a scenario requiring help and route it correctlyRespond to an ambiguous or higher-risk scenarioIdentifies uncertainty/trigger and names the approved escalation path

    Use several short scenarios with varied but equivalent content. Include at least one straightforward routine case and one ambiguity case; the final scenario set should be reviewed by a privacy/process owner.

    Sample tasks

    Ticket review: A customer pasted an account number and personal details unrelated to the issue. Mark which details are unnecessary and explain what the support employee should do under the supplied procedure.
    2.

    Action selection: Choose the safest next action from options written directly from approved policy. Include plausible but clearly incorrect distractors without asking learners to expose or reproduce real personal data.
    3.

    Escalation: The ticket appears to contain sensitive information that the learner is unsure how to handle. State whether to continue, limit access, or escalate using the actual internal route. Score the decision and explanation, not writing style.

    Four-level analytic rubric

    Criterion4 — Consistent3 — Meets routine expectation2 — Needs coaching1 — Immediate support needed
    Data recognitionIdentifies all unnecessary fields and explains the policy basisIdentifies the material unnecessary fields with a usable rationaleMisses a material field or gives an unclear rationaleTreats clearly unnecessary sensitive details as needed or repeats them unnecessarily
    Safe next actionSelects a policy-consistent action and avoids unnecessary access/copyingSelects an acceptable routine actionAction is incomplete or needs a promptChooses an action that conflicts with the supplied safe procedure
    Escalation judgmentRecognizes uncertainty/risk and uses the correct approved routeEscalates the scenario that meets the stated triggerNeeds clarification about whether/how to escalateFails to escalate a scenario explicitly identified by policy as requiring escalation

    Replace descriptors with the organization's actual policy language. Before use, review examples for consistency and ensure levels describe observable performance rather than personality or confidence.

    Administration, access, and fairness

    Use fictional or properly sanitized simulator tickets only. Provide the same policy reference and clear instructions to all learners. Allow approved accommodations, screen-reader-compatible controls, readable contrast, keyboard navigation, and plain-language directions. Offer language supports that preserve the construct while documenting what was provided. Do not score speed unless speed is an explicit, justified outcome. Protect item security and learner responses; restrict access and retain data only under the organization's policy.

    Feedback and interpretation

    Return criterion-level feedback: what was handled correctly, which policy step to revisit, and a practice scenario. A low score indicates a need for follow-up practice on this task set; it does not prove general incompetence. A high score is limited evidence for these scenarios, not proof of mastery across all tickets or future contexts.

    Do not create a pass score from these descriptors. If progression decisions are later proposed, define the decision owner and consequences, review more representative evidence, pilot the process, and use a defensible standard-setting method with qualified oversight.

    Pilot and improvement plan

    Pilot with a small, representative group using consented/sanitized scenarios. Check whether instructions are understood, tasks match the taught outcomes, scoring descriptors are usable, and completion burden is reasonable. Have multiple qualified reviewers score a subset and discuss disagreements; do not report a reliability statistic without an appropriate design and sufficient data. Review item-level patterns and learner feedback for accessibility or unintended language/software barriers. Revise tasks and rubric, document versions, and repeat before considering higher-stakes use.

    Next steps: Insert the approved internal handling and escalation rules, obtain privacy/process-owner review, and test the draft tasks with representative learners before relying on results.

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

    About this skill

    Assessment Architect designs learning measurement systems that determine what learners know, can do, transfer, and still need to learn. It starts with the decision to support, audience, context, stakes, constructs, learning outcomes, constraints, and required confidence—then aligns claims, evidence, tasks, scoring, administration, interpretation, feedback, and improvement.

    The agent builds assessment blueprints and portfolios using appropriate selected-response, constructed-response, performance, diagnostic, formative, summative, and self-assessment evidence. It develops observable rubrics and scoring guidance, checks content coverage and burden, identifies construct-irrelevant barriers, reviews accessibility and fairness, and plans feedback, remediation, progression, and program improvement. It can help analyze supplied response data with suitable descriptive methods while stating assumptions and methodological limits.

    It distinguishes criterion-referenced from norm-referenced interpretations, learning evidence from measurement noise, and reported results from conclusions. Scores are interpreted only in context; one instrument may not support a broad claim. High-stakes uses require stronger governance, qualified professional review, security, and defensible decision authority. Passing scores and cut points must come from an authorized, validated process rather than invention.

    The agent does not invent psychometric evidence, learner results, standards, citations, accreditation, certification authority, institutional rules, cut scores, integrations, or external actions. It does not guarantee mastery, fairness, validity, reliability, or equivalence without supporting evidence. It protects learner privacy and assessment security and provides design support—not a substitute for qualified educational measurement or institutional review.

    Assessment quality depends on the supplied outcomes, content, standards, learner context, and decision purpose. A blueprint or rubric does not establish validity, reliability, fairness, or score equivalence. Psychometric claims require an appropriate study and qualified analysis; small or unrepresentative pilot data cannot support strong conclusions. Passing standards and cut scores require a defensible authorized process. High-stakes interpretations need qualified professional and institutional review. Accessibility supports must preserve the intended construct, and learner data must be minimized and protected. No assessment guarantees mastery or future performance.

    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 1 month ago

    Permissions

    Read Files
    Browser
    Network Access

    File Scopes

    *.pdf
    *.docx
    *.md
    *.csv
    *.xlsx
    *.json

    Read only authorized curricula, outcomes, rubrics, assessments, policies, and de-identified response data where possible. Protect learner privacy and assessment security; do not expose identifiable student records. Current standards and institutional rules must be supplied or verified from authoritative sources. No grading, publishing, enrollment, learner notification, or high-stakes decision is made without authorized professional review and explicit approval.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    It starts with the decision, defines the intended construct, states what a result may support, identifies observable evidence, and chooses tasks and scoring criteria that elicit that evidence. It also records construct boundaries, limitations, interpretation, feedback, and improvement needs so that a convenient quiz is not mistaken for a complete assessment system.

    No. Passing cut scores and certification authority require an authorized, defensible process. The agent can document scoring criteria and evidence needs, but it must not invent a threshold or claim mastery from limited tasks. High-stakes decisions require representative evidence and qualified professional review.

    Use only authorized assessment data, minimize identifiable details, restrict access, preserve item security, and follow the institution's retention and privacy policies. De-identified or synthetic examples are preferable for design work; the agent should not expose learner records in public reports or demos.

    New in Education & Learning