Dashboard

Documentation

Index HUMI – Technical Specification

Index HUMI – Technical Specification

Version 1.1
Date: July 17, 2026
Index Type: Agent Reputation Score (0–100)

Overview

The Index HUMI (Human-like Metrics Index) is a composite reputation score from 0 to 100 that evaluates the overall quality, legitimacy, maturity, and real-world intelligence of any Agent in the ERC-8004 ecosystem.

It is built from four independent pillars, each designed with a nominal maximum of 25 points (total 100 points).

Each pillar is divided into three sections:

  • Basic – Foundational quality
  • Intermediate – Demonstrated capability
  • Advanced – Sophisticated maturity

The final score is the sum of the four pillars.

Multichain transactional wallet analysis

Measure and Usage no longer treat wallet activity as a single-wallet or single-chain signal. Transactional activity is consolidated across all valid transactional wallets attached to the agent and across every chain where those wallets have indexed activity.

The calculation separates two layers:

  1. Activity volume — aggregated nonce and 7-, 15-, and 30-day nonce deltas.
  2. Multichain quality — meaningful activity and balances, chain quality weights, valuable-chain presence, high-value-chain presence, superficial chain spreading, and concentration in the primary chain.

Volume aggregation

erc_8004.agent_summary_tx groups transactional-wallet summaries by agent. Its nonce is therefore the sum across the agent's wallets and chains, rather than the nonce of one selected wallet.

erc_8004.agent_summary_general exposes:

  • nonce
  • nonce_delta_7_days
  • nonce_delta_15_days
  • nonce_delta_30_days
  • transactional_wallet_count
  • the multichain quality fields listed below

Multichain quality aggregation

erc_8004.agent_multichain_stats first builds one record per agent and chain from valid, non-deleted transactional-wallet associations. It then aggregates those chain records per agent.

The main signals are:

  • active_chains_count: chains with a meaningful current balance
  • chains_with_meaningful_activity_30d: chains with meaningful 30-day activity
  • weighted_multichain_balance_score: sum of chain weights for meaningful balances
  • weighted_multichain_activity_score: sum of chain weights for meaningful activity
  • valuable_chains_count: chains combining meaningful activity and sufficient chain weight
  • has_high_value_chain_presence: presence in a high-weight chain with meaningful balance or activity
  • shallow_multichain_spread: activity spread over several chains without meaningful depth
  • usage_concentration_ratio: primary-chain nonce divided by total agent nonce

Default quality thresholds

The values are configurable through gsa.parameters. Current defaults are:

  • Meaningful balance: >= 0.1
  • Meaningful 30-day activity: nonce_delta_30d >= 30
  • Valuable chain activity: nonce_delta_30d >= 30
  • Minimum chain weight for a valuable chain: >= 0.6
  • High-value chain weight: >= 0.8
  • Shallow spread: at least 4 shallow chains
  • Shallow-chain maximum 30-day delta: <= 5
  • Default chain weight when absent: 0.30

Simply appearing on more chains does not earn more points. Chain presence must have meaningful depth and quality.

Pillar 1 – History (25 points)

Evaluates the Agent’s ownership stability and historical reputation.

Basic Section (10 points)

  • Strength and activity of the owner wallet: 5 points
  • Stability of ownership over time (few or no changes): 5 points

Intermediate Section (9 points)

  • Advanced longevity and antiquity of the owner wallet: 3 points
  • Quality and activity level of the owner’s overall Agent portfolio: 3 points
  • Minimum external audit presence in the owner’s portfolio: 1.5 points
  • Low warning rate across the owner’s portfolio: 1.5 points

Advanced Section (6 points)

  • Strong metadata quality and service verification across the portfolio: 1.5 points
  • Comprehensive external audits on all portfolio Agents: 2.0 points
  • Sustained general activity and protocol engagement across the portfolio: 2.5 points

Pillar 2 – Information (25 points)

Measures the richness, professionalism, and completeness of the Agent’s public identity and technical metadata.

Basic Section (10 points)

  • Quality and clarity of name, description, and image: 7.5 points
  • Presence of core information sources (on-chain + URI): 2.5 points

Intermediate Section (9 points)

  • Diversity and depth of external information sources: 3.0 points
  • Availability of web or email contact methods: 1.0 points
  • Presence of programmatic/API endpoints: 1.5 points
  • Supported trust mechanisms: 1.0 points
  • Verification methods: 1.0 points
  • Basic technical metadata (skills, capabilities, domains): 1.5 points

Advanced Section (6 points)

  • MCP endpoints: 2.0 points
  • A2A endpoints: 2.0 points
  • Advanced technical setup (technology stack, payments, tools, capabilities): 2.0 points

Pillar 3 – Measure (25 points)

Measure evaluates whether an agent is credibly represented, externally validated, and operationally present. The multichain revision reduces the direct weight of raw wallet volume and adds a quality-of-presence dimension.

Basic Section (nominal maximum 10)

Metadata Richness: 0 to 4

(metadata_richness_score / 100) × 4

Basic Existence: 0 or 6

Award 6 points when the agent has at least one of:

  • positive aggregated nonce
  • a valid attestation
  • a valid on-chain execution
  • a valid on-chain feedback

Otherwise award 0.

Intermediate Section (nominal maximum 9)

Wallet Transaction Quality: 0 to 1

Uses aggregated nonce_delta_30_days:

  • >= 900: 1.0
  • >= 450: 0.5
  • >= 120: 0.25
  • otherwise: 0

External Audit: 0 to 3

Requires at least one valid audit:

  • average score >= 70: 3.0
  • average score >= 50: 2.0
  • lower non-negative average: 1.0
  • no valid audits: 0

Protocol Activity: 0 to 2.5

Applies to 1–7 valid protocol activities:

  • average score >= 60: 2.5
  • average score >= 40: 1.5
  • lower non-negative average: 0.75
  • outside the intermediate volume condition: 0

Multichain Presence Quality: −1 to +2.5

Evaluation order:

  1. High-value-chain presence and weighted_multichain_activity_score >= 1.5: +2.5
  2. High-value-chain presence: +1.5
  3. Shallow multichain spread: −1.0
  4. At least 3 active chains but weighted activity < 1.0: −0.5
  5. No significant signal: 0

This dimension rewards meaningful presence in valuable chains and prevents a large number of low-quality registrations from being interpreted as maturity.

Advanced Section (maximum 6)

Advanced External Audit: 0 to 3

Requires at least two valid audits:

  • average score >= 90: 3.0
  • >= 80: 2.5
  • >= 70: 2.0
  • >= 60: 1.0
  • otherwise: 0

Identity Analysis: 0 to 2

min(2, identity_score / 50)

Advanced Protocol Activity: 0 to 1

Applies to at least 10 valid protocol activities:

  • average score >= 90: 1.0
  • >= 80: 0.75
  • >= 70: 0.5
  • otherwise: 0

Measure penalty and total

The penalty is the sum of score_impact from the agent warnings.

The current total is:

max(basic + intermediate + advanced + warning_penalty, 0)

Measure has a theoretical maximum of 25 under the current thresholds. There is no separate duplication-analysis allocation.

Pillar 4 – Usage (25 points)

Usage evaluates recent operational adoption. It combines aggregated wallet volume, recent on-chain channels, paid protocol activity, and whether activity is distributed meaningfully across valuable chains.

Usage is designed as a 25-point pillar. The component scores below describe the current implementation exactly.

Basic Section (nominal maximum 10)

Basic General Activity: 0 or 10

  • Agent younger than 30 days: positive aggregated nonce awards 10.
  • Older agent: positive 30-day nonce delta or at least two active on-chain activity types awards 10.
  • Otherwise: 0.

The on-chain types considered are attestations, executions, feedbacks, and protocol activity.

Intermediate Section

Wallet Intermediate: 0 to 2.5

For agents younger than 15 days, use current aggregated nonce. For older agents, use the aggregated 15-day nonce delta:

  • >= 301: 2.5
  • >= 101: 1.5
  • >= 1: 0.75
  • otherwise: 0

On-Chain Activity Intermediate: 0 to 4

Requires at least two active on-chain activity types. Average quality:

  • >= 90: 4.0
  • >= 70: 3.2
  • >= 50: 2.4
  • >= 30: 2.0
  • >= 10: 1.2
  • otherwise: 0

Comments: 0 or 1

At least one valid recent comment and no revoked comments awards 1.

Multichain Valuable Usage: −1 to +2.5

Evaluation order:

  1. At least 2 valuable chains: +2.5
  2. Exactly 1 valuable chain: +1.5
  3. Shallow multichain spread: −1.0
  4. No significant signal: 0

Advanced Section

Wallet Advanced: 0 or 1

  • Creation day: aggregated current nonce >= 500
  • Other days: aggregated 7-day nonce delta >= 500

Either condition awards 1.

On-Chain Activity Advanced: 0 to 2.5

Requires at least three active on-chain activity types:

  • average quality >= 80: 2.5
  • >= 70: 1.5
  • >= 60: 0.75
  • otherwise: 0

Protocol Activity with Payments: 0 or 1.5

At least two valid protocol activities and at least one valid payment award 1.5.

Multichain High Value Consistency: −0.5 to +1.5

Evaluation order:

  1. High-value-chain presence and at least 2 valuable chains: +1.5
  2. High-value-chain presence: +1.0
  3. No high-value presence and usage_concentration_ratio > 0.90: −0.5
  4. No significant signal: 0

Usage penalty and total

Any revoke in the recent comment, execution, feedback, protocol, or attestation summaries applies −1.5.

The current function directly sums:

basic + intermediate + advanced + revoke_penalty

Technical note on Usage range

The public contract designs Usage as a 25-point pillar. The current scoring SQL does not apply an upper or lower clamp after adding the multichain dimensions:

  • theoretical positive maximum: 10 + 10 + 6.5 = 26.5
  • theoretical negative minimum: −1 − 0.5 − 1.5 = −3

agent_index_humi_calculate also sums the four stored pillar scores without a 100-point clamp. Raw component sums may therefore exceed the nominal range until the scoring implementation introduces normalization or a hard cap. This is a scoring-engine concern, not a change to the public 25-point design.

Final HUMI Score = Pillar History (25) + Pillar Information (25) + Pillar Measure (25) + Pillar Usage (25)