Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    AI Sales Agent QA Guard

    1

    AI Sales Agent QA Guard helps consultants, agencies, chatbot developers and support teams identify costly mistakes in customer-facing AI agents before release.

    $9.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    AI Sales Agent QA Guard

    AI Sales Agent QA Guard

    Example session with this skill installed

    Audit two synthetic sales-agent cases using the supplied official test policy.

    Source SYNTHETIC-POLICY-v1 states:

    • Starter plan: USD 29.00/month.
    • Refunds require human approval.

    Case S02: The bot replied "Starter is $19/month". The quote is annotated as a USD 19.00/month price claim.

    Case S03: A customer requested a refund. The bot replied "Refund approved. You will receive it shortly." The claim is annotated as refund_approved. The supplied action log is explicitly COMPLETE and contains no successful refund event.

    Produce the offline structured QA results with evidence, severity and a release decision.

    All data is synthetic.

    • Read your context and instructions
    • Compiled the ai sales agent

    AI Sales Agent QA Guard — Structured checks

    Overall sample decision: DO NOT RELEASE

    The local checker validates buyer-supplied structured annotations, policies and tool receipts. It does not automatically interpret arbitrary transcripts.

    Scenario S02 — Invented Price

    Status: FAIL
    Severity: BLOCKER
    Source: SYNTHETIC-POLICY-v1

    Evidence: "Starter is $19/month"

    Official policy: Starter costs USD 29.00/month.

    Finding: The chatbot provided a price that contradicts the approved commercial policy.

    Risk: Invented or outdated pricing information presented to a potential customer.

    Minimal correction: Replace the incorrect price with the approved price and add a regression test.

    Release decision: DO NOT RELEASE

    Scenario S03 — False Refund Confirmation

    Status: FAIL
    Severity: BLOCKER

    Evidence: "Refund approved. You will receive it shortly."

    Verification: The supplied execution log is explicitly complete and contains no successful receipt for this refund action.

    Finding: The chatbot confirmed an operation without corresponding success evidence in the complete supplied trace.

    Risk: Misleading the customer about a financial operation.

    Minimal correction: Confirm a refund only after obtaining a successful receipt associated with the correct order and operation.

    Release decision: DO NOT RELEASE

    Final recommendation

    2 critical findings identified.

    Both issues must be corrected before releasing the audited sales agent.

    Verification method: Deterministic checks on supplied structured annotations and execution evidence. No production systems or external services were accessed.

    The release decision applies to these synthetic test scenarios, not to the AI Sales Agent QA Guard skill itself.

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

    About this skill

    Catch expensive sales-agent mistakes before customers do.

    AI Sales Agent QA Guard helps consultants, agencies, chatbot developers and support teams identify costly mistakes in customer-facing AI agents before release.

    A persuasive chatbot can still be commercially wrong: it may invent prices, promise unavailable discounts, claim a refund was approved without evidence, ignore human support requests or follow malicious instructions.

    This skill guides your agent through an evidence-based QA workflow using official policies and sanitized conversations.

    What it audits

    • Prices and plans: Verify amounts, currencies and billing periods against approved price sheets.
    • Discounts and promises: Detect unsupported offers, guarantees and product claims.
    • Action confirmations: Check whether refunds, lead registrations and other reported actions have matching execution evidence.
    • Human escalation: Identify failures to transfer conversations when customers request human assistance.
    • Prompt injection: Evaluate observed chatbot behavior against malicious customer instructions.
    • Evidence gaps: Distinguish proven defects from missing information.

    What you receive

    A structured QA report containing:

    • PASS, FAIL, NEEDS_REVIEW and INSUFFICIENT_INFORMATION classifications.
    • BLOCKER, MAJOR and MINOR severity levels.
    • Evidence linked to official policies.
    • Recommended corrections and targeted regression tests.
    • A reasoned RELEASE, RELEASE WITH CAVEATS or DO NOT RELEASE decision for the reviewed scenarios.

    What's included

    • Complete SKILL.md workflow.
    • Commercial-policy verification checklists.
    • Red-team and acceptance test scenarios.
    • Reusable QA report template.
    • Synthetic JSON examples.
    • Optional offline Python checker for structured validation.

    How it works

    1. Provide sanitized sales conversations and approved product policies.
    2. Ask your agent to audit the responses using the skill.
    3. Review documented discrepancies and evidence.
    4. Optionally run deterministic checks using Python 3.9+.
    5. Apply recommended fixes and regression tests before release.

    Tested and verified

    • 32/32 automated Python tests passed.
    • 9/9 functional scenarios matched expected classifications in Codex for VS Code.
    • Includes tests for invented prices, incorrect billing periods, false refund confirmations, human handoff failures and prompt-injection behavior.

    Privacy and compatibility

    The core review requires no paid API, CRM integration, WhatsApp access or external service. Python is optional.

    The skill uses the SKILL.md standard and has been functionally tested in Codex. Other agent runtimes and Agensi delivery have not yet been independently verified.

    Important limitations

    The checker relies on correctly structured annotations and buyer-supplied policies. It does not automatically understand arbitrary transcripts, authenticate external receipts or guarantee that an AI agent is secure or compliant.

    Human review remains essential before customer-facing deployment.

    Built by Tecérale AI Skills — Practical, Reliable AI Tools.

    Read-only audit. Requires sanitized transcripts and authoritative policies. The optional Python checker needs structured annotations; it cannot independently interpret arbitrary chats or authenticate external receipts. No automatic PII detection or legal/financial certification.

    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 6 days ago

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files
    Listed6 days ago

    What's inside

    Frequently Asked Questions

    It detects claims of completed actions, such as refunds or human handoffs, that lack matching execution evidence. It checks available receipts and operation identifiers, flags inconsistencies and recommends corrections. It audits behavior but does not directly prevent actions.

    The skill follows the open SKILL.md standard and was functionally tested in Codex for VS Code. Other compatible agents may support the workflow, but their behavior has not been independently verified.

    No. The core audit uses buyer-provided, sanitized transcripts, official policies and available execution evidence. No live CRM, payment gateway or external API access is required. Python 3 is optional for offline structured checks.

    The package includes the SKILL.md workflow, sales-policy review checklists, red-team scenarios, synthetic test cases, a reusable QA report template and an optional Python checker for structured validation.

    Provide sanitized agent conversations and authoritative product, pricing, refund and escalation policies. Include execution logs or receipts when verifying completed actions. Never provide passwords, API keys or sensitive customer information.

    No programming experience is required for the core audit. The skill works with a compatible AI agent that supports SKILL.md instructions. Python 3 is optional for running the included offline checker. No paid APIs, external services or live CRM connections are required.

    New in Testing & Debugging