Audit report published Aug 2026. Web3 AI agent platform (LLM planner, MCP tools, RAG, on-chain wallet) · Python (examples).
| Severity | Count |
|---|---|
| Critical | 1 |
| High | 4 |
| Medium | 5 |
| Low | 4 |
| Total | 14 |
Safe Edges assessed an autonomous Web3 AI agent platform for a private client. Fourteen issues are listed: one Critical, four High, five Medium and four Low. The first seven cover unvalidated transaction destinations, unverified MCP tool providers, unverified RAG sources, over-privileged wallets, sensitive wallet data in agent memory and logs, and missing rate limits. The remaining seven are the findings of the separate AI and MCP report, kept here in their original position. The original report did not record resolution statuses.
C-01 Prompt Injection Enables Unauthorized On-Chain Transaction ExecutionH-01 Smart Contract Destination Validation Is Not Enforced During Autonomous Transaction ExecutionH-02 MCP Tool Registration Trusts Unverified Tool ProvidersH-03 Arbitrary Smart Contract Interaction via Tool Argument InjectionH-04 MCP Tool Registration Allows Malicious Blockchain ToolsM-01 Retrieval-Augmented Knowledge Base Does Not Verify Source IntegrityM-02 Excessive Wallet Privileges Increase Autonomous Execution RiskM-03 RAG Poisoning Influences Autonomous Trading DecisionsM-04 Excessive Wallet Permissions Increase Blast RadiusM-05 Cross-Agent Trust Boundary Allows Strategy ManipulationL-01 Agent Memory Retains Sensitive Wallet Context Across SessionsL-02 AI Tool Invocation Metadata Is Excessively LoggedL-03 Missing Request Rate Limiting for Autonomous Agent OperationsL-04 Sensitive Wallet Metadata Persisted in Long-Term Agent MemorySend the repository and a commit hash through the contact form, message @bugtester25 on Telegram, or book a 30-minute scoping call. 200+ protocols audited · $4B+ secured · 0 hacks post-audit. Prefer email? info@safeedges.in.