Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    UX3 Product Design Harness

    1

    Review product decisions across users, evidence and business. Get a clear verdict and the next bounded action.

    Free

    1 useSecurity scanned
    UX3 Product Design Harness

    UX3 Product Design Harness

    Example session with this skill installed

    Illustrative worked fixture, not a customer session. Use UX3 Product Design Harness to review the small onboarding copy change described in the included resources/examples/quick-gate-review.json fixture. Apply the supplied evidence and scope. Return the complete canonical review and validate it with resources/scripts/check_review.py. Do not invent customer outcomes.

    • Read your context and instructions
    • Compiled the ux3 product design

    Included Quick Gate worked fixture

    This is the complete shipped illustrative example, not customer evidence or measured conversion data. Its canonical validation passed in this session and in GitHub CI. Validation checks structure and internal consistency, not evidence truth.

    {
      "schema_version": "2.1",
      "session_config": {
        "working_language": "en",
        "canonical_identifiers": "en",
        "fallback_language": "en"
      },
      "review_id": "example-quick-001",
      "product_direction": "Clarify the empty-state action on the setup page.",
      "mode": "quick_gate",
      "stage": "prototype",
      "actor_boundary": {
        "target_population": [
          "New workspace admins"
        ],
        "operator": "Workspace admin",
        "beneficiary": "Workspace team",
        "machine_actor": "Not applicable",
        "affected_people": [
          "Invited team members"
        ],
        "excluded_groups": [
          "Existing configured workspaces"
        ]
      },
      "product_organization": {
        "organization": "Example Product Co.",
        "project_owner": "Activation team",
        "issuer": "Example Product Co."
      },
      "stakeholder_effects": [
        {
          "stakeholder": "Invited team members",
          "effect": "Clearer setup can reduce delayed invitations."
        }
      ],
      "human_judgment_boundary": {
        "human_owned_decisions": [
          "Final copy tone"
        ],
        "machine_supported_inputs": [
          "Support notes",
          "Session replays"
        ],
        "not_mechanized": [
          "Taste and brand judgment"
        ],
        "approval_gates": [
          {
            "gate_id": "product-owner-release-approval",
            "accountable_owner": "Product owner",
            "condition": "Before release to users."
          }
        ]
      },
      "unresolved_assumptions": [
        "Whether copy or missing product guidance causes the hesitation."
      ],
      "combined_verdict": "continue",
      "weakest_flow": "evidence",
      "strongest_signal": "Support notes and session replays show repeated hesitation at the same empty state.",
      "riskiest_claim": "Clearer copy will improve first-action completion.",
      "main_uncertainty": "Whether copy or missing product guidance causes the hesitation.",
      "confidence_level": "medium",
      "evidence_tier": "t2",
      "main_trade_off": "Limit the change to copy so the team can learn without expanding onboarding scope.",
      "human_owned_decision": "none",
      "human_decision_record_id": null,
      "evidence_receipt_ids": [
        "support-pattern-001",
        "session-replay-pattern-001"
      ],
      "reviews": {
        "user": {
          "verdict": "continue",
          "strongest_signal": "People repeatedly pause before the first setup action.",
          "weakest_assumption": "The action label is the primary source of confusion.",
          "missing_evidence": "Post-change completion behavior.",
          "disconfirming_evidence": "People still pause after the copy change.",
          "stop_condition": "The change reduces comprehension in a five-person check.",
          "evidence_tier": "t2"
        },
        "evidence": {
          "verdict": "continue",
          "strongest_signal": "Two independent internal sources show the same hesitation.",
          "weakest_assumption": "The replay sample represents new users.",
          "missing_evidence": "Segmented first-action completion.",
          "disconfirming_evidence": "Completion does not move after release.",
          "stop_condition": "No measurable event can distinguish the old and new copy.",
          "evidence_tier": "t2"
        },
        "business": {
          "verdict": "continue",
          "strongest_signal": "The change is reversible and may improve activation.",
          "weakest_assumption": "First-action completion predicts later value.",
          "missing_evidence": "Activation-to-retention link.",
          "disconfirming_evidence": "Completion rises while retained setup does not.",
          "stop_condition": "The change expands beyond copy and one action label.",
          "evidence_tier": "t2"
        }
      },
      "next_smallest_action": "Release one copy variant and compare first-action completion.",
      "execution_boundary": {
        "may_do": [
          "Edit the empty-state headline and one action label."
        ],
        "must_not_do": [
          "Change layout, permissions, data handling, or onboarding steps."
        ],
        "must_ask": [
          "A material scope, risk, data, or stakeholder impact changes."
        ]
      },
      "human_approval_gates": [
        {
          "gate_id": "product-owner-release-approval",
          "accountable_owner": "Product owner",
          "condition": "Before release to users."
        }
      ],
      "high_risk_actions": [],
      "stop_conditions": [
        "People still hesitate after the copy change.",
        "The change requires onboarding scope expansion."
      ],
      "context_pack_required": true
    }
    

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

    About this skill

    Move from a product idea to a decision your team can inspect and your agent can follow.

    UX3 Product Design Harness adds structured product judgment to the AI agent you already use. Product leaders, founders, and designers supply a decision and its evidence. The agent reviews who benefits, what the evidence supports, and whether the value exchange can remain viable.

    Every review ends with one of three verdicts: continue within a stated boundary, verify one specific proof step, or stop and reframe the question. The weakest flow determines the next action. A strong business pitch cannot cancel out missing user evidence.

    Use it before committing to a feature, designing an experiment, reviewing launch readiness, or turning user feedback into the next iteration. Start with Quick Gate for a small reversible decision, Standard Gate for a feature or workflow, and UX3 Council when the decision needs deeper review.

    The complete skill includes prompts, schemas, decision rules, templates, validators, and worked examples. Validation checks output structure and internal consistency. It cannot prove that supplied evidence is true or replace accountable human judgment.

    Install the free MIT-licensed skill from the canonical GitHub repository. To discuss adoption in your team's product workflow, connect with the author: https://www.linkedin.com/in/cis2042/.

    Source and complete resources: https://github.com/cis2042/ux3-product-design-harness Install: npx skills add cis2042/ux3-product-design-harness --skill product-design-harness Website: https://cis2042.github.io/ux3-product-design-harness/

    Validators check structure and internal consistency, not whether supplied evidence is true. The included reviews are illustrative examples, not customer results. Human-owned decisions remain accountable to people. Requires an existing AI agent; Python 3 and jsonschema for validation. Routing is advisory and cannot bypass canonical review.

    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 25 days ago

    • Free with an Agensi account

    Requires

    Python 3
    jsonschema
    Python
    Node.js

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Reads installed resources and supplied product evidence, and writes bounded review outputs. Local validation uses Python and jsonschema. No external account, API key or network is required for the included validators. Optional external evidence research remains subject to the operator’s permissions.

    Listed25 days ago

    What's inside

    Frequently Asked Questions

    Review User Flow, Evidence Flow and Business Flow together. The worst lane verdict and deterministic weakest flow control the next action; stronger lanes cannot cancel out missing evidence. The result states a bounded action and stop conditions. This is a review contract, not a guarantee of business success.

    The MIT-licensed package includes SKILL.md, LICENSE and the complete resources directory: prompts, ontology and rules, schemas, templates, Python validators, documentation and worked examples. Install the entire folder; SKILL.md alone is not the complete runtime.

    Use an agent that supports the open SKILL.md format, can read the adjacent resources and can run the required validators. The skills CLI install path is verified in GitHub CI. Runtime-specific integration may vary; the listing does not certify every agent shown by the marketplace.

    Use Python 3 and the bundled jsonschema dependency requirements. No hosted UX3 service, external account or additional paid API is required for local validation. Your existing AI agent and any optional external research may have separate provider costs.

    Record accountable human decisions and designer taste explicitly. They are not converted into machine scores. Validators check structure, required controls and internal consistency; they cannot establish that evidence is true or replace human judgment.

    New in Business & Operations