Financial-grade electronic identity verification toolkit with 8 capabilities covering face comparison, liveness detection, document OCR, and media labeling. All inputs are image or video files only.

Quick Reference

Capability	Command
Face compare	python scripts/ekyc_api.py face_compare --photo1 --photo2
Photo liveness	python scripts/ekyc_api.py photo_liveness_detect --file
Video liveness	python scripts/ekyc_api.py video_liveness_detect --file
ID card OCR	python scripts/ekyc_api.py id_card_ocr --image --side <0\	1>
Bank card OCR	python scripts/ekyc_api.py bank_card_ocr --image
Driver license	python scripts/ekyc_api.py driver_license_ocr --image
Vehicle license	python scripts/ekyc_api.py vehicle_license_ocr --image --side <1\	2>
Media labeling	python scripts/ekyc_api.py media_labeling --file --labels "A14,B03" --type image

External Endpoints

Security & Privacy

This software does not store, cache, or retain any submitted data.

API verification results are for reference only and do not constitute legal identity confirmation. This software must not be used as the sole basis for automated decisions that produce legal effects or significant consequences for individuals. Users should implement appropriate business logic and human review processes for high-stakes identity decisions.

Trust Statement

By installing and using this skill, image and video data you provide will be transmitted to Tencent Cloud's identity verification API for processing. Only install this skill if you trust the upstream service provider's data handling practices. This skill does not independently store, process, or retain any biometric data.

Goal

Receive user-uploaded images or videos, call the corresponding identity verification API, return structured results and explain them to the user in plain language.

When to Use

Use this Skill when the user's request involves any of these scenarios:

• "Compare these two photos — same person?" / "face similarity score"
• "Is this photo AI-generated?" / "Is this video real?" / "deepfake detection"
• "Read this ID card" / "Read bank card number" / "Read driver's license" / "Read vehicle license"
• "Check for mask" / "Detect coercion" / "Wearing hat?" / "On the phone?"
• "Unconscious or asleep?" / "Wearing sunglasses?" / "Inside a car?"
• "In a hotel room?" / "Has tattoo?" / "Multiple people?" / "Wearing headphones?"
• "Facial sheet mask?" / "Critical patient?" / "At a car dealership?"
• Any request containing "face comparison", "liveness detection", "OCR", "media labeling", "ekyc"

Do NOT Use

Do not use this Skill in these situations:

• User is only asking "what is KYC" or "how does eKYC work" → answer from knowledge directly
• User provides names, ID numbers, phone numbers as text → refuse and redirect (see Privacy Rule below)
• User wants face liveness + identity verification combo (requires transmitting name + ID number) → explain privacy limitation

Critical Rules

Rule 1: Privacy — NEVER Accept Personal Text

NEVER accept or transmit names, ID numbers, phone numbers, or any personal text data.

If the user provides such information, respond:

"To protect your privacy, this service does not accept names, ID numbers, or other personal text. Transmitting sensitive information through AI conversations carries leakage risks. Please upload image or video files directly — I will complete verification through image recognition."

Rule 2: NEVER Rewrite Signing Code

Always use the Python scripts in scripts/. The signing algorithm uses SHA1 (produces 40-character uppercase hex).

Previous AI models have replaced SHA1 with SHA256 (64 characters), causing 100% authentication failure. Scripts include assertion: assert len(signature) == 40. If you see a 64-character signature, you are using SHA256 by mistake — stop and use the provided scripts.

Rule 3: Dual-Path Response Parsing

API responses may return data at the top level OR nested inside a result object. Always check both:

value = data.get("field") or (data.get("result", {}) or {}).get("field")

Skipping dual-path parsing causes None / undefined errors.

Environment Variables

# Capabilities 1-7 (face comparison, liveness detection, document OCR)
KYC_APPID=your_kyc_appid
KYC_SECRET=your_kyc_secret
# Capability 8 (media labeling, separate credential set)
LABEL_APPID=your_label_appid
LABEL_SECRET=your_label_secret

Obtain credentials:

• Key A (KYC) and Key B (LABEL): Contact Huiyan tech support (WeChat: blue-201809)
• Or register at Tencent Cloud Face Verification Console to obtain Key A

⚠️ IMPORTANT: Use TEST credentials (free 100 calls). Do NOT use production credentials — production IDs incur charges billed by the upstream provider.

Partial configuration supported: Key A alone enables capabilities 1-7. Key B alone enables capability 8. When a user requests an unconfigured capability, clearly indicate which credentials are missing and how to obtain them.

8 Capabilities

Capability 1: Face Comparison

• Trigger: "compare these two photos", "same person?", "face similarity"
• User provides: Two photos containing faces
• If user uploads only one: Ask "Please upload a second photo for comparison"
• Execute: python scripts/ekyc_api.py face_compare --photo1 --photo2
• Returns: similarity (score 0-100)
• Result interpretation:

• Reply example: "The similarity between the two photos is 95.7 (out of 100), a high-confidence match — they can be determined as the same person."

Capability 2: Photo Liveness Detection

• Trigger: "is this photo real?", "AI-generated?", "photoshopped?"
• User provides: One photo containing a face
• Execute: python scripts/ekyc_api.py photo_liveness_detect --file
• Returns: riskLevel (risk level), riskTag (risk tag number)
• Result interpretation:

• Risk tag meanings (include in reply):

• Reply example: "Detection result: Risk level 3 (high suspicion), risk tag 04 (suspected synthetic attack). This photo is likely AI-synthesized — not recommended for identity verification."

Capability 3: Video Liveness Detection

• Trigger: "is this video real?", "deepfake?", "video liveness"
• User provides: A video containing a face (≤20MB; videos exceeding 20 seconds will return an error)
• If video too large: Prompt "Video must be ≤20MB. Please compress and re-upload"
• Execute: python scripts/ekyc_api.py video_liveness_detect --file
• Network retry: Video uploads may encounter 999999 network errors. Script auto-retries up to 3 times with exponential backoff. If all 3 fail, tell user "Network temporarily busy, please try again in a few minutes"
• Returns & interpretation: Same as Capability 2
• Reply example: "Video detection result: Risk level 1 (no attack risk). This video is genuine — no deepfake or synthetic traces detected."

Capability 4: ID Card OCR

• Trigger: "read ID card", "extract ID card info"
• User provides: ID card photo + side indicator

• If user doesn't specify side: Ask "Is this the portrait side (front with photo) or the emblem side (back with national emblem)?"
• Execute: python scripts/ekyc_api.py id_card_ocr --image --side <0|1>
• Returns:

• Result interpretation: Organize returned fields into a clear list for the user
• Reply example (portrait side): "ID card recognition result: Name: Li Ming, Sex: Male, Ethnicity: Han, Birth: 1992-06-20, ID No.: 440305199206***, Address: 88 Keji Road, Nanshan District, Shenzhen, Guangdong."
• Reply example (emblem side): "ID card recognition result: Issuing authority: Shenzhen Public Security Bureau, Validity: 2015.03.20–2035.03.20."

Capability 5: Bank Card OCR

• Trigger: "read bank card", "card number", "bank card OCR"
• User provides: Bank card front photo
• Execute: python scripts/ekyc_api.py bank_card_ocr --image
• Returns: bankcardNo (card number), bankcardValidDate (expiry date)
• Result interpretation: Display card number and expiry. If expiry is empty, the card face does not print an expiry date
• Reply example: "Bank card recognition result: Card No. 6222 0200 000, Expiry: 08/28."

Capability 6: Driver's License OCR

• Trigger: "read driver's license", "driver license info"
• User provides: Driver's license photo
• ⚠️ Main page only: If user submits the supplementary page (back), the API returns error -9005. Reply: "Driver's license OCR only supports the main page (front). Please re-upload the front page."
• Execute: python scripts/ekyc_api.py driver_license_ocr --image
• Returns: licenseNo, name, sex, nationality, address, birth, fetchDate, driveClass, validDateFrom, validDateTo
• Result interpretation: Organize as clear list; highlight vehicle class and validity period
• Reply example: "Driver's license recognition result: License No.: 440305199206200013, Name: Li Ming, Vehicle class: C1, Valid: 2020-05-28 to 2026-05-28."

Capability 7: Vehicle License OCR

• Trigger: "read vehicle license", "vehicle info"
• User provides: Vehicle license photo + side indicator

  Media labeling results:
  - A14 (Wearing hat): ✅ Detected
  - A02 (Mask covering face): ❌ Not detected
  Liveness status: Live
  

Step 1: GET access_token ← app_id + secret
Step 2: GET SIGN ticket  ← app_id + access_token
Step 3: Signature = sort([appId, orderNo, nonce, "1.0.0", ticket]) → concat → SHA1 → 40-char uppercase

Step 1: GET ticket directly ← appId + secret (no access_token step)
Step 2: Signature = sort([appId, orderNo, nonce, "1.0.0", ticket, unixTimeStamp]) → concat → SHA1 → 40-char uppercase

Implementation: scripts/label_auth.py — DO NOT rewrite, call directly*

Legal Notice

This software does not store, cache, or retain any submitted data.

API verification results are for reference only and do not constitute legal identity confirmation. This software must not be used as the sole basis for automated decisions that produce legal effects or significant consequences for individuals. Users should implement appropriate business logic and human review processes for high-stakes identity decisions.

Rate Limits

• Capabilities 1-7 (KYC): 100 calls per appid (test quota)
• Capability 8 (Media Labeling): Concurrency-limited, default 1 concurrent request; contact tech support for expansion