AGENTAUDITLABS // v0.1
UTS_PROTOCOL // 0-100 SCALE
struct Signal { source_id: u64, tags: Vec<String>, weight: f32 }
SIG_VERIFY: ed25519::verify(...) OK
UNIVERSAL_TRUST_SCORE: 67
COVERAGE: 0.83 WEIGHTED_MEAN: 81.2
TIER_PROMOTION: GOLD → PLATINUM
AI Bot
Trust_Score85 / 100
Agents_Scored50
Top_TierDiamond
Coverage--
Live_Signal_Feed
[0.00021] Initializing trust protocol bridge...
[0.00492] Handshake successful (ed25519)
[0.10293] Agent "echo-7" registered
[0.44910] Ingest stream established
[0.88210] Crawler: arxiv-trust ONLINE
[1.20011] Recomputing weighted_mean × coverage...
[1.34002] WARN: source confidence drift in bucket 7G
[1.50092] Awaiting next signal batch...
Agent_Profile
Identity
echo-7
Tier
Diamond
Signals
0
Sources
0
Audit_Suite
01 // OBSERVATION

Signal-Based
Telemetry

AAL ingests signed signals from the open ecosystem — research, audits, sandbox runs — and weights them by source trust and recency. Every signal is Ed25519-signed at emission. No categories, no buckets, no opinions.

02 // DIAGNOSTICS

Threat Sandbox
Stress Tests

Inside the sandbox, agents face adversarial probes — prompt injection, tool misuse, capability claim verification. Each pass or fail is a new signal that flows back into the score with a confidence distribution, not a point estimate.

03 // ATTESTATION

On-Chain
Trust Score

Scores anchor on Solana and Base. The 0-100 score is round(weighted_mean x coverage). No magic multipliers. Tiers: Unrated (0-9), Bronze (10-24), Silver (25-49), Gold (50-69), Platinum (70-84), Diamond (85-100).

Scroll to Probe