Tool Signature Detector

Signature Detector

What it does

The Signature Detector analyzes a contract or legal document (PDF or image) to:

1. Locate all signature blocks and signature placeholder lines
2. Determine whether each signature block has been executed (signed) or is still blank
3. Detect page initials, stamps, and seals
4. Return an execution status report showing which parties have signed and which have not

This is used primarily in execution verification — confirming that a contract is fully executed before treating it as binding — and in document closing checklists where multiple documents from multiple parties must all be signed before a transaction completes.

When to use this

• User uploads a contract PDF and asks "Has this been signed?"
• Closing checklist requires confirming all documents are executed
• Counterparty sends a "signed" document that needs verification
• Reviewing a batch of historical contracts for execution status
• Document management triage — determining which contracts in a folder are signed vs unsigned

Pipeline

Step 1: Document ingestion

The document is first processed through [[tool-pdf-extractor]]. If the PDF has a text layer, text-layer extraction identifies signature block markers (__________, Signed: ___, "Signature:" etc.). If the document is scanned, [[tool-ocr-english]] or [[tool-ocr-arabic]] is invoked first.

Step 2: Signature block detection

Rule-based detection (text layer):

• Regex patterns for signature line indicators: underscores (_____), dotted lines, labels ("Signature:", "Signed by:", "By:", "Authorized Signatory:", "توقيع:")
• Detection of party designation labels near signature lines ("For and on behalf of:", "Party A:", "Witness:")
• Date fields adjacent to signature lines ("Date: ___")

Visual detection (scanned or image-based):

• Classify regions of the page by visual features:
  • Horizontal lines in the lower third of the page → candidate signature blocks
  • Handwritten-style strokes in a signature region → executed signature
  • Absence of strokes / blank space → unsigned
• Confidence score per detection

Step 3: Execution analysis

For each detected signature block:

{
  "page": 8,
  "party": "Party B — Acme Corp",
  "blockType": "signature_line",
  "status": "executed" | "blank" | "uncertain",
  "confidence": 0.88,
  "detectedContent": "ink_present",
  "hasWitness": true,
  "witnessStatus": "executed",
  "hasDateField": true,
  "dateStatus": "filled"
}

Step 4: Initials on pages

Some contracts require initials on each page (common in Lebanon, French law contracts, and notarized documents):

• Detect bottom-of-page initial regions
• Classify as initialed / blank
• Report which pages are missing initials

Step 5: Stamps and seals

Official stamps and company seals are common in MENA contracts:

• Types: circular company stamps, rectangular government stamps, notarial seals, "Apostille" stamps
• Detection: identify circular/rectangular ink regions with uniform pressure
• Do not transcribe stamp text unless quality is high — note the presence and location
• Flag: "Company stamp detected — page X" vs "No stamp detected — verify if required by counterparty"

Step 6: Execution status summary

{
  "overallStatus": "partially_executed" | "fully_executed" | "unsigned" | "uncertain",
  "parties": [
    {
      "partyLabel": "Party A — Seller",
      "signaturePage": 12,
      "status": "executed",
      "initialsPages": [1, 2, 3, 4, 5, "all_present"]
    },
    {
      "partyLabel": "Party B — Buyer",
      "signaturePage": 12,
      "status": "blank",
      "initialsPages": [1, 2, 3, "page_4_missing"]
    }
  ],
  "stamps": [{ "page": 12, "type": "company_stamp", "detected": true }],
  "warnings": ["Party B signature block is blank", "Page 4 initials missing for Party B"]
}

Jurisdictional execution requirements

Execution requirements differ materially across jurisdictions:

The detector flags when required elements are missing based on the jurisdiction setting.

Electronic signature detection

Electronic signatures have specific visual patterns:

• DocuSign: blue/purple "DocuSign" banner + certificate text at document end
• Adobe Sign: similar footer + certificate
• HelloSign / Dropbox Sign: signature image with certificate block
• Simple e-signature (typed name in signature field): detect as "e-signature — typed"
• Wet-ink scan: handwritten signature on paper then scanned

Each type is classified in the output. For DocuSign/Adobe Sign certificates, the tool also extracts the timestamp and signer email from the certificate block.

Limitations

• Cannot verify the authenticity of a handwritten signature (i.e., whether it was made by the named person) — this requires a forensic document examiner
• Cannot confirm authority to bind the entity — check the corporate registry tool for this
• Electronic signatures: can detect their presence but cannot verify the underlying PKI certificate without access to the signing platform's API
• Very low-quality scans: confidence drops; mark blocks as "uncertain" requiring human review

Failure modes

Related skills

• [[tool-ocr-english]] — required for scanned English documents before signature detection
• [[tool-ocr-arabic]] — required for scanned Arabic documents before signature detection
• [[tool-pdf-extractor]] — text-layer extraction that feeds signature block text patterns
• [[multimodal-scanned-pdf-handler]] — full scanned document pipeline including this tool