Loading ToolBench
Solana threat intelligence system. Tracks serial rug-pull operators, bot clusters, and funding networks using persistent operator memory. Provides mint-level verdicts for operator risk, token risk, and rug pull detection via mainnet monitoring.
SolSentry defines 4 tools with clear, domain-specific descriptions (avg 150 chars) and proper JSON schemas. All tools are read-only threat intelligence queries with well-structured inputs. However, parameter descriptions are minimal (wallet_address, mint_address lack format/validation guidance), output schemas are undocumented (LLMs cannot predict response structure), and error handling lacks recovery guidance. Tool naming is strong (verb_noun pattern: check_*, get_*, explain_*). No tool annotations (readOnlyHint) despite all being read-only. Server-side validation via requireStringArgs() is good defensive practice but doesn't compensate for missing output documentation.
Check the risk profile of a Solana wallet as a token operator/deployer. Returns confirmed rug count, risk label, behavioral tags, and a plain-English summary. Call this before interacting with tokens deployed by an address.
Check the risk profile of a Solana token by mint address. Returns risk score, risk flags, operator history, and bundle detection. Call this before buying or listing any token.
Get a human-readable explanation of why a wallet or token is risky. Accepts either a wallet address (checked as operator) or a token mint address. Returns a plain-English warning suitable for displaying to end users.
Get SolSentry system-wide threat intelligence statistics: total scans, accuracy, resolve rate, operator counts, bot clusters.
Output schemas undocumented. LLMs cannot predict response structure (risk score format, operator history fields, bundle detection fields). Forces agents to guess at downstream field names.
Parameter descriptions lack format/validation guidance. 'wallet_address' and 'mint_address' don't specify Base58 encoding, length, or validation rules. LLMs may pass invalid addresses.
No tool annotations. All 4 tools are read-only but lack readOnlyHint annotation. Agents cannot distinguish safe tools from destructive ones without reading descriptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
Error responses lack recovery guidance. SolSentryError returns status + body but no actionable next steps. 'API 404' tells agent nothing; should suggest 'Address not found. Verify Base58 encoding or try check_network_stats() for available data.'
explain_risk accepts ambiguous 'address' parameter (wallet OR mint). No guidance on how LLM should distinguish them. Parameter description should clarify: 'Solana address (wallet or token mint). Tool auto-detects type.'