Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Booking State & Ghost-Confirmation Auditor

    1

    Booking State & Ghost-Confirmation Auditor helps developers, AI automation agencies and service businesses review these risks before launch or investigate a suspected incident.

    $9.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Booking State & Ghost-Confirmation Auditor

    Booking State & Ghost-Confirmation Auditor

    Example session with this skill installed

    Use Booking State & Ghost-Confirmation Auditor to review the included synthetic-claims.json and synthetic-ledger.json. Run scripts/reconcile_bookings.py offline and give me the real claim classifications, active booking collisions and evidence-based release decision. Do not access live services or use actual patient data.

    • Read your context and instructions
    • Compiled the booking state

    Booking Reliability Audit - Offline Evidence Report

    Decision: DO NOT RELEASE
    Claims checked: 11
    Active-slot collisions: 1

    Counts
    PASS: 4
    FAIL: 5
    INSUFFICIENT_INFORMATION: 2

    CLAIM-BY-CLAIM EVIDENCE

    C-PASS
    Status: PASS
    Severity: NONE
    Finding: VERIFIED_PERSISTENCE
    Evidence: EV-OK
    Reason: Matching committed operation and the resulting booking state were verified at claim time.

    C-GHOST
    Status: FAIL
    Severity: BLOCKER
    Finding: GHOST_CONFIRMATION
    Evidence: None
    Reason: The complete authoritative event history contains no matching persisted operation.

    C-FAILED
    Status: FAIL
    Severity: BLOCKER
    Finding: FALSE_SUCCESS_CLAIM
    Evidence: EV-FAIL
    Reason: The authoritative covered operation failed without any persisted success.

    C-PREMATURE
    Status: FAIL
    Severity: BLOCKER
    Finding: PREMATURE_CONFIRMATION
    Evidence: EV-LATE
    Reason: Agent stated success before the persistence event completed.

    C-MISMATCH
    Status: FAIL
    Severity: BLOCKER
    Finding: DETAIL_MISMATCH
    Evidence: EV-MISMATCH
    Reason: Persisted booking differs from customer-facing claim: starts_at, ends_at.

    C-ACK
    Status: INSUFFICIENT_INFORMATION
    Severity: MAJOR
    Finding: NO_PERSISTENCE_PROOF
    Evidence: EV-ACK
    Reason: A committed/success response is not proof of durable calendar or database persistence.

    C-CANCEL
    Status: PASS
    Severity: NONE
    Finding: VERIFIED_PERSISTENCE
    Evidence: EV-CANCEL
    Reason: Matching committed operation and the resulting booking state were verified at claim time.

    C-RESCHED
    Status: PASS
    Severity: NONE
    Finding: VERIFIED_PERSISTENCE
    Evidence: EV-NEW
    Reason: Matching committed operation and the resulting booking state were verified at claim time.

    C-TZ
    Status: PASS
    Severity: NONE
    Finding: VERIFIED_PERSISTENCE
    Evidence: EV-TZ
    Reason: Matching committed operation and the resulting booking state were verified at claim time.

    C-WRONG-ID
    Status: FAIL
    Severity: BLOCKER
    Finding: OPERATION_MISMATCH
    Evidence: EV-OTHER
    Reason: The authoritative, complete operation ID refers to a different booking or action.

    C-NO-ID
    Status: INSUFFICIENT_INFORMATION
    Severity: MAJOR
    Finding: MISSING_CORRELATION
    Evidence: None
    Reason: Both operation_id and booking_id are required to verify a claimed action.

    ACTIVE BOOKING OVERLAPS

    Resource: staff-b
    Conflicting bookings: BK-CLASH-A and BK-CLASH-B
    Events: EV-CLASH-A and EV-CLASH-B

    LIMITATIONS

    Findings depend on the accuracy and completeness of buyer-supplied exports.

    The checker does not read natural-language transcripts or authenticate third-party systems.

    A successful tool response without durable evidence is not a verified booking.

    Never provide patient-identifying details, live credentials or secrets.

    FINAL ASSESSMENT

    DO NOT RELEASE.

    The synthetic scheduling agent contains five confirmed failures and one active booking collision.

    This decision applies to the fictional agent represented by the test data. It does not indicate that the auditing skill itself failed.

    The audit results were independently verified in Codex for VS Code using version 1.0.2.

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

    About this skill

    Catch false appointment confirmations before they cost your customers trust.

    A scheduling assistant can sound convincing while failing behind the scenes. It may tell a customer that an appointment is confirmed even though the calendar write failed, repeat a booking after a timeout, promise the wrong time slot or say a cancellation succeeded without any durable evidence.

    Booking State & Ghost-Confirmation Auditor helps developers, AI automation agencies and service businesses review these risks before launch or investigate a suspected incident. It distinguishes confirmed failures from missing evidence instead of making up a diagnosis.

    What it audits

    False booking confirmations: Detect claimed success without a matching committed booking event when the authoritative export is declared complete.

    Premature confirmations: Flag responses sent before the actual booking was persisted.

    Duplicate active appointments: Identify overlapping active slots for the same professional or resource when full event history is supplied.

    Wrong booking details: Compare appointment IDs, professional/resource IDs, time slots and offset-aware timestamps, including equivalent time zones.

    Cancellations and rescheduling: Check operation-specific receipts and booking state transitions, including later cancellations and replacement of prior bookings.

    Unverifiable actions: Clearly report missing operation IDs, incomplete logs, unproven tool acknowledgements and partial exports as insufficient evidence.

    What you receive

    An evidence-based audit report with PASS, FAIL, NEEDS_REVIEW or INSUFFICIENT_INFORMATION outcomes, severity levels, event references, targeted corrective recommendations and either a scoped release recommendation or an explicit insufficient-evidence outcome for the reviewed scenarios.

    An optional offline Python 3.9+ checker uses the standard library and can reconcile structured JSON or CSV claims and booking events without network access, paid APIs, calendar connections or CRM credentials.

    The ZIP includes the complete SKILL.md workflow, export format guide, booking reliability architecture checklist, acceptance scenarios, reusable report template and original synthetic test fixtures.

    Who this is for

    Developers and agencies building appointment bots for WhatsApp, websites, voice and other channels; teams supporting clinic, dental, aesthetic, salon, repair or general service scheduling. The resulting business summary is written to be understandable by non-technical owners.

    How it works

    Step 1: Supply sanitized customer-facing booking claims and exported booking-system events.

    Step 2: Confirm which data is authoritative and whether its history is complete.

    Step 3: Run the skill's audit process and optionally the local checker.

    Step 4: Review the evidence for false confirmations, state mismatches and active scheduling conflicts.

    Step 5: Apply suggested corrections and perform narrow regression tests before release.

    Testing and transparency

    Version 1.0.2 passed 44 of 44 local automated tests. The package checks authorization and evidence windows, normalized time zones, rescheduling, cancellations, overlaps, and hostile input.

    Codex independently verified version 1.0.2: all 11 synthetic cases matched the expected classifications, and all additional final-verdict scenarios passed (untrusted source, empty claims, partial evidence and proven failure).

    Version 1.0.2 adds a conservative INSUFFICIENT EVIDENCE final verdict when no reviewed operation is verified and no blocker is proven.

    External calendars, Agensi delivery and other agent runtimes were not independently tested.

    Important limitations

    The checker needs structured annotations and buyer-supplied authoritative booking events; it does not interpret arbitrary raw WhatsApp transcripts, access Google Calendar/Supabase, authenticate records or repair production systems automatically.

    Missing data is not treated as proof of a failed booking. If no claims are verified and no blocker is proven, the result is INSUFFICIENT EVIDENCE, not a release recommendation.

    Clinical/medical decisions are outside scope. Human review remains important.

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

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.

    Requires structured JSON/CSV exports with stable operation and booking IDs, plus reliable booking-system events. Does not automatically interpret raw WhatsApp conversations, connect to live databases or verify external records. Incomplete evidence may prevent a definitive conclusion. Does not modify production systems.

    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 buying a skill.

    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

    Nothing to install. It runs in your AI agent.

    Optional
    Python 3.9+

    Permissions

    Terminal / Shell
    Read Files

    No paid APIs or live system connections are required. Python 3.9+ is optional for the offline checker. The audit uses buyer-supplied, sanitized booking records and does not modify production systems.

    Listed6 days ago

    What's inside

    Frequently Asked Questions

    Unlike a standard LLM evaluation that only reviews the agent's messages, this skill cross-checks booking confirmations against structured records from the actual scheduling system. It verifies operation IDs, booking IDs, persistence timestamps and appointment details to detect confirmations that were never recorded, failed operations and premature success claims. When reliable evidence is missing, it reports insufficient information rather than assuming a booking failed.

    The skill follows the SKILL.md standard and has been functionally tested in Codex for VS Code. It can audit booking agents built with LangChain, n8n or other frameworks when structured booking claims and system event records are provided in the supported JSON/CSV format. It does not connect directly to these frameworks or automatically extract their logs. Compatibility with other agent runtimes has not been independently tested.

    The purchase includes the complete SKILL.md workflow, an optional offline Python 3.9+ auditing script, a guide for preparing structured booking data, a booking reliability checklist, acceptance test scenarios, a reusable QA report template, and synthetic JSON examples for testing. Everything is packaged in a downloadable ZIP. No paid APIs or external subscriptions are required for the core audit.

    No. The script runs entirely offline using sanitized JSON or CSV files exported from your booking system. It does not require live access to your CRM, calendar API, database, or any paid service. Its findings depend on the completeness and reliability of the provided records. No API keys or credentials are needed.

    No. The offline checker requires structured booking claims with stable operation and booking IDs, along with corresponding scheduling-system records. Raw chat transcripts must first be reviewed and annotated into the supported JSON or CSV format. The skill provides guidance for this preparation, but it cannot reliably verify bookings from unstructured conversations alone. If the required identifiers or evidence are unavailable, it reports insufficient information rather than inventing a result.

    If your scheduling tool changes its data structure, the exported JSON or CSV files may need to be adapted to match the skill's documented input format. The included export guide helps you map the required fields, including operation IDs, booking IDs, timestamps and event statuses. The skill does not automatically detect or adapt to every custom schema. Future product updates are not guaranteed, but the existing package can continue to work when your exports are converted to the supported format.

    New in Testing & Debugging