Loading ToolBench
Evaluate trustworthiness of any domain, URL, wallet, or API. Returns 0-100 score with 5 sub-scores: SSL, DNS, WHOIS, security headers, content. Zero-cost, zero-dependency. The trust layer agents need before interacting with unknown services.
Two tools with well-structured schemas and comprehensive descriptions. trust_score_evaluate has excellent parameter documentation (target, checks array with enum constraints) and detailed output explanation covering 5 sub-scores. trust_score_batch_compare similarly well-defined with minItems/maxItems constraints. Descriptions are 200-400 chars, within production baseline (p10=34, p90=392). Both tools include clear use-case guidance and anti-patterns (what NOT to use them for). However, output schemas are described in prose rather than formal JSON Schema, and no explicit error handling guidance is documented. Tool names follow verb_noun pattern (evaluate, compare) appropriately.
Use this when you need to compare the trustworthiness of multiple domains, URLs, or wallets and pick the safest option. Accepts 2-5 targets and returns trust scores for all, sorted from most to least trusted. Returns: { mostTrusted: "...", leastTrusted: "...", results: [{ compositeScore, grade, verdict, subscores }] } Use cases: - Comparing API providers before choosing one - Verifying multiple domains from a search result - Ranking wallet addresses by on-chain reputation - Due diligence on a list of services Do NOT use for single targets -- use trust_score_evaluate instead (cheaper at $0.01 vs $0.02).
Use this when you need to check if a domain, website, API endpoint, or crypto wallet is safe to interact with. Returns a composite trust score 0-100 with letter grade (A+ to F), verdict (trusted/moderate/suspicious/dangerous), and 5 sub-scores: 1. SSL/TLS (25%): certificate validity, HSTS, expiry, issuer 2. DNS (15%): A/AAAA/MX/NS records, SPF, DMARC, DNSSEC 3. WHOIS (25%): domain age, registrar reputation, expiry, suspicious TLDs 4. Security Headers (20%): CSP, X-Frame-Options, HSTS, Referrer-Policy, Permissions-Policy 5. Content (15%): reachability, latency, status code, robots.txt, CORS For wallet addresses (0x...): returns on-chain trust based on transaction count, ETH/USDC balance, contract status on Base L2. Example output: { compositeScore: 82, grade: "A", verdict: "trusted", subscores: { ssl: { score: 90, ... }, dns: { score: 85, ... }, ... } } Use this BEFORE making payments, sending sensitive data, or trusting any external service. Essential for agent safety. Do NOT use for SEO analysis -- use seo_audit_page instead. Do NOT use for email validation -- use email_verify_address instead. Do NOT use for tech stack detection -- use website_detect_tech_stack instead. Do NOT use for port scanning -- use network_scan_ports instead.
Output schemas documented in prose only, not formal JSON Schema. Descriptions state 'Returns: { compositeScore, grade, verdict, subscores }' but no structured schema definition visible in code.
No error handling guidance. Descriptions lack recovery hints (e.g., 'If domain unreachable, try with a different endpoint' or 'Timeout after 8s, retry with subset checks'). LLM has no guidance on retryability or next steps on failure.
trust_score_evaluate 'checks' parameter uses enum but lacks constraint documentation in description. Description says 'Default: all' but does not state that passing empty array is invalid or what happens if an invalid check name is passed.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | 2025-06-18+ | v2 |
No pagination or result limits documented. trust_score_batch_compare accepts 2-5 targets (good), but trust_score_evaluate has no stated limit on result size or latency bounds. Description mentions '3-8s' for full scan but no timeout guarantee or partial-result fallback.
Descriptions include anti-patterns (what NOT to use tools for) but lack positive use-case prioritization. 'Do NOT use for SEO analysis' is helpful, but no explicit guidance on when to prefer trust_score_evaluate vs trust_score_batch_compare beyond cost ($0.01 vs $0.02).