Trust Score Calibration

TRUST SCORE // FORENSIC CALIBRATION

The Sentinacle Trust Score is a penalty-based algorithm returning a value from 0 to 95, displayed on a scale of 100, summarizing the security posture of a token based on direct on-chain evidence.

Base score: 95. Penalties are subtracted for each finding detected. Bonuses are applied for verified positive signals (clean simulation, confirmed liquidity, contextual recovery for protocol tokens). The score is clamped between 0 and 95.

A Trust Score of 95/100 means: simulation confirmed the token is not a honeypot, taxes are 0%, no significant risk signals were detected, and contextual signals confirm legitimate protocol design. It does not mean the project is sound as an investment.


Score Labels

ScoreLabelMeaning
80 – 95🟢 TRUSTEDNo critical or high-risk signals detected. Clean simulation confirmed.
50 – 79🟡 MODERATESome risk signals present but no confirmed exploit. Review findings carefully.
20 – 49🟠 RISKYSignificant risk signals. Multiple high or critical findings present. High caution.
10 – 19🔴 HIGH RISKCritical vulnerability confirmed — honeypot, sell impossible, or catastrophic signal.
0 – 9⛔ CRITICALConfirmed direct theft mechanism. Do not interact.

Signal Sources & Their Weight

Sentinacle applies different penalty weights depending on the source of the finding. Not all signals carry the same evidentiary value.


Tier 1 — Direct On-Chain Evidence (Highest weight)

motor_propio — Sentinacle’s own simulation and bytecode engine

The most authoritative source. Findings are derived from live transactions on a fork of the blockchain (EVM) or direct RPC reads (Solana).

SeverityPenalty
CRITICAL (honeypot confirmed, sell impossible)−100
HIGH (dangerous opcode or authority confirmed)−50
MEDIUM (bytecode pattern — not a confirmed exploit)−12
LOW−5

holders / solana_holders — Holder concentration

Computed directly from on-chain token accounts. Statistically validated predictor of rug pulls (p < 0.001 in academic literature). For Solana, program-owned vaults (PDAs) are excluded — only individual wallet concentration is penalized.

SeverityEVM PenaltySolana Penalty
CRITICAL (top 10 individual wallets > 80%)−55−60
HIGH (top 10 > 50%)−28−20
MEDIUM (top 10 > 30%)−5−5

lp_analysis — Liquidity pool

Single wallet controlling most of the pool is a Tier 1 rug pull signal.

SeverityPenalty
CRITICAL (> 80% of LP in single unknown wallet)−60
HIGH−32
MEDIUM−8
LOW (locked in verifiable contract)0
SUCCESS (LP burned)+5

Tier 2 — Contextual On-Chain Signals

governance — Governance & control analysis

SeverityPenalty
CRITICAL (hidden owner, exploitable upgrade path)−45
HIGH (modifiable fees, dangerous admin function, EOA proxy admin)−22
MEDIUM (unrenounced ownership, unidentified controller)−8
LOW (Gnosis Safe without timelock)−3
SUCCESS (Gnosis Safe + Timelock, or renounced)+3

behavioral — On-chain behavioral patterns

SeverityPenalty
CRITICAL−30
HIGH (fresh deployer, suspicious pattern)−15
MEDIUM−5

market — Confirmed market data

SeverityPenalty
CRITICAL (confirmed $0 liquidity)−40
HIGH (extreme liquidity mismatch)−35

Tier 3 — External Sources (Reduced weight)

legacy — External APIs and third-party data (GoPlus cross-check)

Used only as a secondary validation source — never as primary evidence. Penalties are reduced due to known false-positive rates.

SeverityPenalty
CRITICAL−15
HIGH−8
MEDIUM−3

Known false positives from external sources are fully ignored:

  • Proxy / Delegate contracts
  • Mint function / Mintable
  • Blacklist / Blacklisted
  • Pausable / Transfer Paused
  • Anonymous Team
  • Unverified source code (handled separately as a fixed −12 penalty)

Bonuses

Simulation Bonus (applied once)

ConditionBonus
EVM: clean Anvil simulation (not honeypot, sell tax < 50%)+20
Solana: active Raydium pool confirmed, no sell restriction+5

Liquidity Bonus

Confirmed on-chain liquidity signals an active market and reduces the relative weight of unconfirmed risk signals.

On-chain liquidityBonus
≥ $5,000,000+20
≥ $500,000+12
≥ $50,000+5

Reputation Floor

Tokens on Sentinacle’s verified reputation list (major stablecoins, Tier-1 protocol tokens) have a minimum Trust Score floor applied regardless of findings.


Context Modifier

Before bonuses are applied, the engine evaluates the token’s category classification and applies contextual score recovery for findings that are expected design features in legitimate protocols — not scam signals.

How it works

Every token is classified into one of four categories based on verifiable on-chain signals:

CategoryDescription
Cat. 1Unknown issuer. Full penalties apply. No contextual recovery.
Cat. 2Established market presence (community token, narrative token). Partial recovery.
Cat. 3Verifiable protocol (team, governance, deployer history). Significant recovery.
Cat. 4Infrastructure asset (reputation list). Reputation floor applies.

What gets contextual recovery

FindingCat. 2 mediumCat. 2 highCat. 3 mediumCat. 3 high
Mint Authority Active (HIGH or CRITICAL)+10+20+30+45
Freeze Authority Active (HIGH or CRITICAL)+10+20+30+45
Ownership Not Renounced+5+10
Concentrated holders (protocol vaults)+30+50

What never receives recovery

Honeypot confirmed, sell tax > 50%, single anonymous wallet > 30% of LP, single whale > 30% of supply. These represent direct theft vectors regardless of protocol reputation.

Classification confidence

Confidence is determined by total classification points accumulated:

PointsConfidence
≥ 10High
7 – 9Medium
6Low

Recovery amounts scale with confidence — a Cat. 3 high token receives more recovery than a Cat. 3 medium token.


Simulation Veto

When Sentinacle’s own simulation confirms not a honeypot and sell tax < 50%, penalties from external (legacy) sources are automatically halved. Direct on-chain simulation evidence supersedes inferred signals from third-party APIs.

Additionally, motor_propio HIGH and MEDIUM findings (bytecode patterns, not confirmed exploits) are also halved when a clean simulation contradicts them — the live execution is considered more authoritative than static analysis.


Inconclusive Simulation Handling

If a buy transaction reverts during simulation (due to anti-bot protection, timestamp gate, or low liquidity), the engine does not declare CRITICAL unilaterally. The result is logged as LOW “Simulation Inconclusive” and GoPlus is queried as a cross-check. A CRITICAL honeypot finding is only generated when both sources independently agree.


Worked Example: JLP (Jupiter Perps LP)

StepValueNotes
Base score95
SimulationClean (0% tax, sell confirmed)
Holder concentration100% in program vaultsPDAs excluded — not penalized
Mint / Freeze authorityBoth renounced+SUCCESS
ClassificationCat. 2 high (liquidity $2.4B, “perps” LP name pattern, sim clean)
Context recovery0No contextualizable findings present
Sim bonus+5 (Solana)
Liquidity bonus+20 ($2.4B)
Final score95 TRUSTED

Representative Score Ranges

Token profileTypical rangePrimary driver
Major DeFi LP token (JLP, mSOL)90–95Protocol vaults excluded; clean sim; high liquidity
Governance token with VC vesting80–92Some holder concentration (real); clean sim
New EVM token, source verified, Gnosis Safe75–90Protocol signals present; context recovery applies
Established memecoin (BONK, PEPE)70–85Cat. 2; some concentration; sim clean
New token, mint active, EOA owner50–70Mint HIGH; no context recovery (Cat. 1)
High concentration, no simulation20–50CRITICAL holders; missing sim bonus
Confirmed honeypot0CRITICAL motor_propio: −100

⚠️ A high Trust Score indicates technical security, not market safety. It means the contract code cannot steal your funds through code manipulation. It does not protect against whale selling, poor tokenomics, or market volatility. Always perform your own due diligence (DYOR).