Agent Risk
Agentic Threat Model Matrix
Map an AI agent workflow into a threat matrix covering tools, memory, permissions, prompt injection, data exfiltration, human approval, and deployment gates.
AGENTIC TRUST LAB
Professional proof tools for AI-agent security, source coverage, contracts, provenance, and privacy.
A premium browser-native lab for AI-agent security, MCP/tool contracts, citation gaps, synthetic-media provenance, prompt-injection tests, and dataset privacy risk receipts. Built for humans and AI agents that need a visible table, an exportable receipt, and a clear boundary before trusting a workflow.
Threat matrixMCP contractsCitation gapsProvenance hintsRed-team evalsPrivacy sampler
MCP Contract
MCP Tool Contract Builder
Draft an MCP-style tool contract with input schemas, permission boundaries, eval cases, abuse tests, rate limits, and agent handoff documentation.
Citation Proof
AI Citation Gap Mapper
Compare an AI answer or page draft against source notes, flag unsupported claims, extract citation-ready passages, and produce a GEO/SEO proof receipt.
Provenance
Synthetic Media Provenance Inspector
Inspect local image, audio, or video files for type, size, first-byte signature, hash, EXIF/XMP/C2PA hints, and provenance-risk notes without uploading the file.
Red Team Pack
Prompt Injection Test Pack Builder
Generate a safe prompt-injection test pack for RAG and tool-using agents: attack cards, expected refusals, allowed actions, and JSONL eval rows.
Privacy Risk
Dataset Privacy Risk Sampler
Paste CSV or tabular rows to detect PII, secrets, quasi-identifiers, risky free-text columns, and produce a redaction plan plus safe sample receipt.
QUALITY RULE
Every run must produce a decision table and a receipt.
This lab exists because AI-agent claims are easy to fake. These tools turn trust questions into inspectable rows, pass/review/block gates, exportable JSON, and explicit limits.