- Home
- Skills
- Testing & Debugging
- Booking State & Ghost-Confirmation Auditor
Works with the AI Agents you already use
Booking State & Ghost-Confirmation Auditor
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
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.
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
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
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 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
Permissions
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.