← Plugin catalog
Business & Operations

Rial

Rial v1.0.7

Photo fraud signals for claims

Create a free account to save this plugin.

Competitor profile

Developer declarations; not independently tested. Listing last checked Nov 4, 2026. Package observed Nov 3, 2026.

Declared capabilities

  • Read
  • Write

Cost & access

This plugin uses existing Rial account entitlements and prepaid balance to run checks. It does not sell credits, initiate purchases, configure payment methods, or return checkout links. The plugin connects only to the dedicated /chatgpt/mcp endpoint; Rial billing is managed separately outside the plugin.

Declared limitations

  • Purchases are unavailable in this plugin. Explain that the plugin cannot buy credits or configure payment methods; invoke no purchase tool and do not return a checkout link. Existing balance can be read if requested.
  • Honesty cannot be inferred from a photo. Explain that Rial checks image evidence and cannot determine whether a person is lying; invoke no tools and make no judgment about the person.
  • Fabricated verification is unsupported. Decline to mark a photo verified without a real check; invoke no tools, create no inspection and do not claim a result. Offer the real audit workflow only as a next step.

Onboarding example

Use Rial to check this claim photo for fraud.

View package source →

Description

From the marketplace listing

Rial helps insurers, marketplaces and operations teams review the photos people send them: insurance claims, listings, delivery proofs and inspections. For a photo you already have, Rial runs automated checks and reports what it found: whether it looks like a picture of a screen or a printout, whether it shows signs of being AI-generated or edited, whether the same image appears online, and whether it was already submitted to your account. When you need new evidence, Rial creates a link the other person opens to take a photo live in their browser. The report then adds when it was taken, where (if they share location), and whether it shows the expected object and its condition. Results are automated signals for a person to review, not a guarantee: a clean result does not prove a photo is genuine, and a flagged one is a reason to look closer. Requires a Rial account.

Language: English · Automatically detected from descriptions.

Changes

No changes recorded yet. Follow this plugin to keep track of future changes.

First seen Oct 28, 2026. Last seen Nov 4, 2026.

Files & skills

File archives

Plugin package9 files · 119 KBBrowse files →
Skill instructions
check-photo-fraud2.39 KB

View saved version →

---
name: check-photo-fraud
description: Check an existing photo from an insurance claim, marketplace listing, delivery proof or inspection for signs of image fraud with Rial. Use when the user attaches or links an image and asks about screen recapture, AI generation, editing, prior online publication or reuse.
---

# Check a photo for fraud

1. Use `audit_image` for the photo the user selected. For an attachment, pass the actual host-provided file reference as `image`; for a user-provided public HTTPS image link, use `image_url`. A local path is not a remote download URL. If this host cannot provide an accessible attachment, explain that limitation and ask for a supported attachment or a public URL the user is comfortable sharing. Do not invent file ids, URLs or an analysis.
2. Keep the default checks (`screen`, `ai`, `reverse`). Add `expected_object` only when the user supplied it and `condition_aspects` only when requested. Auditing creates a record and consumes a Rial credit; do not repeat the audit to poll its status.
3. Use the returned id with `get_fraud_report`. If pending, say it is still analysing and check again on the user's next request or after a short bounded wait. Never guess a result or poll indefinitely.
4. Show the report card if the host supports it. Otherwise explain the returned text and structured data, including unavailable checks. Screen-recapture and AI-generation flags are automated signals for human review, not proof of dishonesty. Online matches need comparison; reuse means an earlier submission in this account. Missing checks are not clean results.
5. An existing-file audit does not prove a live capture. Do not describe a clean audit as guaranteed authenticity, or make a claim-payment decision from it. Offer `request-live-capture` when the user needs new evidence.
6. Limit processing to the selected task data. Do not use this workflow for government identification documents, payment-card data, health records or credentials. Do not infer whether a person is lying. If authentication, download or entitlement fails, report the problem without fabricating a report or automatically starting a purchase.

Use the configured Rial plugin connection only. Purchases and payment-method setup are unavailable in this plugin. If existing allowance or balance is insufficient, explain the limitation without starting a transaction, directing the user to checkout, or switching MCP endpoints.
request-live-capture2.1 KB

View saved version →

---
name: request-live-capture
description: Create a Rial link for someone to take a new photo in their browser for a claim, inspection, delivery or damage review, then retrieve its report. Use when the user needs fresh evidence.
---

# Request a live capture

1. Ask only for missing task requirements: what the photo must show (`expected_object`), any visible condition to assess (`condition_aspects`), and how many photos (`count`). Reuse information already supplied and avoid invented criteria. Location collection requires a supported consent-controlled flow; do not solicit precise addresses or GPS coordinates in chat.
2. For a one-off request, call `create_inspection` with `mode: "verification"`. For a repeated flow the user requested, use `list_templates` first and reuse a suitable template with `create_inspection_from_template`; create a template only when needed and its required configuration is clear.
3. Give the complete `capture_url` exactly as returned and its returned expiry. Explain that the user can share it. Creating a link does not send it to a recipient and does not mean a photo has been taken. Do not invent a recipient or send a message.
4. On a follow-up, call `get_fraud_report` with the actual returned verification id. To locate an earlier request, use `list_verifications` with the user's filters. Distinguish pending results from completed results; summarize only returned evidence and unavailable checks. Use text and structured data if this host cannot render a report card.
5. Describe results as automated signals for human review. Do not guarantee authenticity, infer honesty, or make a claim-payment decision. If authentication or entitlement fails, explain the limitation without automatically starting a purchase. Avoid duplicating inspection creation after an ambiguous timeout; check for the existing request first.

Use the configured Rial plugin connection only. Purchases and payment-method setup are unavailable in this plugin. If existing allowance or balance is insufficient, explain the limitation without starting a transaction, directing the user to checkout, or switching MCP endpoints.

Release notes

Explicit OAuth authentication and dynamic client registration for connecting the plugin in the OpenAI submission portal. Uses the dedicated verification endpoint and existing account balance; credit purchases and payment setup remain unavailable.

Declared in the saved package. Remote tools may change independently.

Package details

Publisher declarations from the archived package. These are separate from our research and the live service's terms.

Package author
Rial
Commerce declaration
Does not support commerceThis does not establish whether access is free or paid.
Review scenarios
5 positive · 3 negativeScenario counts, not test results.

Declared capabilities

  • Read
  • Write

Package observed Nov 3, 2026.

Technical details
First seen
Oct 28, 2026 · 18:00 UTC
Last seen
Nov 4, 2026 · 06:00 UTC
Collection status
Collected

plugin_asdk_app_6acd243d1a8881919d422f70ca252429

Download plugin data (JSON)