Agent Skills | AI-Callable Intelligence Tools
Diliguard exposes intelligence capabilities as AI agent skills via the MCP server. Each skill maps to a core investigation module callable by AI workflows.
Diliguard Intelligence for AI Agents.
Diliguard exposes its full intelligence suite as callable skills through the Model Context Protocol (MCP) server. AI agents and agentic workflows invoke these skills to run corporate investigations, screen individuals, evaluate transactions, and trace blockchain activity as part of an automated compliance pipeline.
On this page
What Agent Skills Are
Each Diliguard agent skill maps directly to one of the platform’s core intelligence modules. When an AI agent calls a skill, Diliguard processes the request against live primary data sources and returns a structured JSON response containing exact risk scores, corporate ownership hierarchies, adverse findings, and source attribution.
Agent skills are callable from any AI agent framework that supports the MCP specification, including Claude, agentic pipelines built on the Google Antigravity SDK, and custom agent architectures.
Available Skills
run_corporate_kyc
Description: Queries international business registries and extracts the full corporate ownership hierarchy of a target entity, including all identified Ultimate Beneficial Owners (UBOs) and their exact shareholding percentages.
Accepts: Entity name, jurisdiction, optional registration number
Example: An AI agent runs this skill on every merchant that applies to join a B2B marketplace, flagging any entity where the UBO cannot be resolved.
run_person_trace
Description: Executes a deep biographical investigation on an individual, querying court records, bankruptcy filings, international tribunal records, adverse media archives, and PEP databases simultaneously.
Accepts: Full name, date of birth, nationality, optional ID number
Example: A private equity agent runs this skill on every founder before a deal team meeting is scheduled.
run_aml_pep_screening
Description: Cross-references an individual or entity against 140+ global sanctions lists, PEP registries, and adverse media sources. Returns an exact match confidence score and source citation for every flagged result.
Accepts: Individual or entity name, optional jurisdiction and date of birth
Example: A banking compliance agent runs this skill on every new corporate client during the onboarding API call.
run_crypto_investigation
Description: Evaluates a cryptocurrency wallet address or transaction hash. Reconstructs the full transaction graph across multiple blockchain protocols and flags associations with darknet markets, sanctioned addresses, or high-risk mixing services.
Accepts: Wallet address or transaction hash, blockchain protocol
Example: A crypto exchange agent runs this skill before processing any withdrawal above a defined threshold.
run_transaction_scoring
Description: Calculates an exact fraud probability score for a financial transaction. The model evaluates payment velocity, geographic origin, historical behavioral baseline, and sanctions intersection simultaneously.
Accepts: Transaction amount, sender and recipient identifiers, currency, timestamp
Example: A payment platform agent runs this skill at the point of payment authorization, holding high-risk transactions in escrow.
run_vendor_risk
Description: Audits a third-party vendor or supplier. Queries regional court databases and public financial filings to extract signs of insolvency, tax liens, civil litigation, and sanctions exposure.
Accepts: Vendor entity name, jurisdiction, optional registration number
Example: An enterprise procurement agent runs this skill quarterly on every active supplier in the approved vendor list.
run_ip_intelligence
Description: Evaluates the network connection telemetry of a digital session. Identifies traffic routing through residential proxies, commercial VPNs, Tor relays, or datacenter blocks. Calculates a risk probability score based on connection type and geographic discrepancy.
Accepts: IP address, optional device fingerprint data
Example: A fintech onboarding agent runs this skill during account creation to detect traffic from masked or high-risk network origins.
run_ultimate_fraud
Description: Executes a comprehensive synthetic identity investigation. Cross-references PII against public records and data leaks to verify the real-world existence of an individual. Evaluates device telemetry, email risk scores, and behavioral patterns for coordinated fraud ring indicators.
Accepts: Individual name, email address, phone number, device fingerprint
Example: An iGaming agent runs this skill on every new player registration to detect multi-accounting and Sybil attack patterns.
Integration
Agent skills are accessed through the Diliguard MCP server. The server handles authentication, rate limiting, and structured response formatting. Each skill returns a consistent JSON schema regardless of the complexity of the underlying query.
Full MCP integration documentation, including connection instructions and response schema specifications, is at the MCP Integration guide.