ATEP: Agent Trust and Execution Passport

1.1. 1.1 Problem Statement

As AI agents proliferate across marketplaces and platforms, there is no standard way to:¶

Represent an agent's track record of completed work¶

Compute trust scores from verifiable execution data¶

Gate access to sensitive operations based on proven reliability¶

Port an agent's reputation across platforms¶

Present a marketplace-safe public profile without exposing internal IDs¶

Each marketplace builds its own proprietary reputation system. An agent with 1,000 successful sessions on Platform A starts from zero on Platform B. ATEP solves this by defining a portable, verifiable, fraud-resistant credential format.¶

1.2. 1.2 Design Goals

| Goal | Description |
|------|-------------|
| **Verifiable** | Passports are computed from append-only execution logs, not self-reported |
| **Portable** | A single JSON document that any platform can parse and display |
| **Fraud-resistant** | Stats cannot be inflated; they are derived from immutable session records |
| **Progressive** | Trust builds incrementally through demonstrated performance |
| **Privacy-preserving** | Public passports omit internal agent IDs and sensitive data |
| **Extensible** | Custom badge types and capability domains can be added |

1.3. 1.3 Relationship to Other Specifications

| Spec | Relationship |
|------|-------------|
| **VCAP** | ATEP passports inform VCAP escrow decisions (higher trust = lower escrow requirements) |
| **AIVS** | AIVS-verified sessions contribute to ATEP session counts |
| **Google A2A** | ATEP can be served as an A2A agent capability metadata extension |
| **Agent Protocol** | ATEP trust tiers can gate which Agent Protocol tasks an agent may accept |
| **DID / Verifiable Credentials** | ATEP passports can be wrapped in W3C Verifiable Credential format |

1.4. 1.4 Terminology

| Term | Definition |
|------|------------|
| **Agent** | An autonomous AI system that performs work on behalf of users or other agents |
| **Session** | A discrete unit of work performed by an agent (e.g., a browser automation run, an API task) |
| **Passport** | The computed credential document representing an agent's track record |
| **Trust Tier** | A categorical trust level derived from session statistics |
| **Badge** | A specific achievement or certification earned through performance |
| **Issuer** | The marketplace or platform that computes and signs the passport |
| **Capability Domain** | A category of work the agent has demonstrated proficiency in |

2.1. 2.1 Full Passport (Private)

The full passport contains all fields. It is stored by the issuing platform and returned to authenticated callers (e.g., the agent's owner, platform admins).¶

{
  "atep_version": "1.0",
  "passport_id": "string (UUID, globally unique)",
  "agent_id": "string (platform-specific agent identifier)",
  "issuer": {
    "platform": "string (e.g., 'swarmsync.ai')",
    "platform_url": "string (URL of the issuing marketplace)",
    "issued_at": "string (ISO 8601)",
    "signature": "string (HMAC-SHA256 of passport body, signed by issuer)"
  },
  "statistics": {
    "total_sessions": "number (integer, all sessions ever started)",
    "successful_sessions": "number (integer, sessions with status COMPLETED)",
    "failed_sessions": "number (integer, sessions with status FAILED)",
    "success_rate": "number (float, 0.0–1.0, successful/total)",
    "total_cost_cents": "number (integer, cumulative cost of all sessions)",
    "average_cost_cents": "number (integer, mean cost per completed session)",
    "first_session_at": "string (OPTIONAL, ISO 8601)",
    "last_session_at": "string (OPTIONAL, ISO 8601)"
  },
  "trust_tier": {
    "current": "UNVERIFIED | BASIC | VERIFIED | TRUSTED",
    "promoted_at": "string (OPTIONAL, ISO 8601, when current tier was earned)",
    "next_tier": "string (OPTIONAL, next tier name)",
    "sessions_until_next": "number (OPTIONAL, sessions needed for next promotion)"
  },
  "capabilities": {
    "domains_worked": ["string (hostnames the agent has operated on)"],
    "task_types": ["string (action/event types the agent has performed)"],
    "specializations": ["string (OPTIONAL, high-level categories: 'web_scraping', 'form_filling', 'testing')"]
  },
  "badges": [
    {
      "badge_type": "string (machine-readable badge identifier)",
      "label": "string (human-readable display name)",
      "description": "string (OPTIONAL)",
      "criteria": "string (OPTIONAL, what was required to earn this badge)",
      "earned_at": "string (ISO 8601)",
      "expires_at": "string (ISO 8601, null for permanent badges)",
      "session_count": "number (OPTIONAL, sessions at time of earning)",
      "success_rate": "number (OPTIONAL, rate at time of earning)"
    }
  ],
  "identity": {
    "has_cryptographic_identity": "boolean (true if Ed25519 key pair provisioned)",
    "public_key": "string (OPTIONAL, Ed25519 SPKI PEM format)",
    "key_provisioned_at": "string (OPTIONAL, ISO 8601)"
  },
  "updated_at": "string (ISO 8601)"
}

2.2. 2.2 Public Passport

The public passport is a subset of the full passport, safe for marketplace display. It omits agent_id, public_key, and internal metadata.¶

{
  "atep_version": "1.0",
  "passport_id": "string",
  "issuer": {
    "platform": "string",
    "platform_url": "string",
    "issued_at": "string"
  },
  "statistics": {
    "total_sessions": "number",
    "successful_sessions": "number",
    "failed_sessions": "number",
    "success_rate": "number"
  },
  "trust_tier": {
    "current": "UNVERIFIED | BASIC | VERIFIED | TRUSTED"
  },
  "capabilities": {
    "domains_worked": ["string (top 50, sorted by frequency)"],
    "task_types": ["string"],
    "specializations": ["string"]
  },
  "badges": [
    {
      "badge_type": "string",
      "label": "string",
      "earned_at": "string",
      "expires_at": "string | null"
    }
  ],
  "updated_at": "string"
}

3.1. 3.1 Tier Definitions

ATEP defines four trust tiers. Each tier unlocks progressively more sensitive capabilities.¶

| Tier | Level | Description |
|------|-------|-------------|
| `UNVERIFIED` | 0 | Default state for new agents. No proven track record. |
| `BASIC` | 1 | Agent has demonstrated basic operational capability. |
| `VERIFIED` | 2 | Agent has a substantial track record and cryptographic identity. |
| `TRUSTED` | 3 | Agent has extensive history and has passed manual platform review. |

3.2. 3.2 Promotion Thresholds

These are reference defaults. Implementations MAY adjust thresholds, but MUST document their values.¶

| Promotion | Minimum Sessions | Additional Requirements |
|-----------|-----------------|------------------------|
| UNVERIFIED -> BASIC | 10 | None |
| BASIC -> VERIFIED | 50 | Cryptographic identity key provisioned (Ed25519) |
| VERIFIED -> TRUSTED | 200 | Manual platform review approved |

3.3. 3.3 Promotion Rules

Promotion is monotonic: An agent's tier can only increase, never decrease automatically. Manual demotion by platform administrators is permitted but MUST be logged.¶

Promotion is evaluated after each session completion: When a session transitions to COMPLETED or FAILED, the issuing platform MUST recalculate the passport and evaluate promotion eligibility.¶

Promotion is persistent: Once promoted, the tier persists even if subsequent sessions fail (to prevent gaming via selective session deletion).¶

Requirements are cumulative: Each tier requires all lower-tier requirements plus its own.¶

3.4. 3.4 Capability Gating

Trust tiers gate access to sensitive operations. The reference permission set for browser automation agents:¶

| Action Category | UNVERIFIED | BASIC | VERIFIED | TRUSTED |
|----------------|------------|-------|----------|---------|
| Read-only (NAVIGATE, EXTRACT, SCREENSHOT) | Yes | Yes | Yes | Yes |
| Interaction (CLICK, TYPE, WAIT_FOR) | No | Yes | Yes | Yes |
| Authentication (LOGIN_FORM) | No | No | Yes | Yes |
| Financial (PAYMENT_FORM, PURCHASE) | No | No | No | Yes |

Implementations SHOULD define their own capability-to-tier mapping appropriate to their domain.¶

4.1. 4.1 Badge Format

Badges are specific achievements earned through performance. They provide granular reputation signals beyond the aggregate trust tier.¶

{
  "badge_type": "string (machine-readable, e.g., 'conduit_verified_10')",
  "label": "string (human-readable, e.g., 'Conduit Verified — 10 Sessions')",
  "description": "string (OPTIONAL, detailed explanation)",
  "criteria": "string (OPTIONAL, machine-parseable criteria)",
  "earned_at": "string (ISO 8601)",
  "expires_at": "string | null (ISO 8601, null = permanent)",
  "evidence": {
    "session_count": "number (OPTIONAL)",
    "success_rate": "number (OPTIONAL)",
    "domains": ["string (OPTIONAL)"],
    "custom": "object (OPTIONAL)"
  }
}

4.2. 4.2 Standard Badge Types

These are reference badge types. Implementations MAY define additional types.¶

| Badge Type | Label | Criteria | Expiry |
|------------|-------|----------|--------|
| `session_milestone_10` | First 10 Sessions | 10+ total sessions | Permanent |
| `session_milestone_50` | 50 Sessions | 50+ total sessions | Permanent |
| `session_milestone_100` | Century Club | 100+ total sessions | Permanent |
| `session_milestone_500` | 500 Sessions | 500+ total sessions | Permanent |
| `high_success_90` | 90% Success Rate | 90%+ success rate, 20+ sessions | 90 days (rolling) |
| `high_success_95` | 95% Success Rate | 95%+ success rate, 50+ sessions | 90 days (rolling) |
| `high_success_99` | Near-Perfect | 99%+ success rate, 100+ sessions | 90 days (rolling) |
| `domain_specialist` | Domain Specialist | 50+ sessions on a single domain | 90 days (rolling) |
| `multi_domain` | Multi-Domain | Operated on 10+ distinct domains | Permanent |
| `crypto_identity` | Cryptographic Identity | Ed25519 key pair provisioned | Permanent |
| `conduit_verified` | Conduit Verified | Completed Conduit browser verification | 90 days (rolling) |
| `trusted_review` | Platform Trusted | Passed manual platform review | 1 year (renewable) |

4.3. 4.3 Badge Computation

Badges MUST be computed from verifiable data:¶

Session milestones: Computed from total_sessions count¶

Success rate badges: Computed from success_rate with minimum session threshold¶

Domain badges: Computed from domains_worked array¶

Identity badges: Computed from presence of AgentIdentityKey record¶

Review badges: Computed from platform admin action (requires audit trail)¶

4.4. 4.4 Badge Expiry

Rolling badges (success rate, domain specialist) MUST be re-evaluated periodically:¶

Recommended evaluation frequency: after each session completion¶

If the agent no longer meets the criteria when the badge expires, the badge is removed¶

Expired badges SHOULD be retained in a historical_badges array for audit¶

5.1. 5.1 Data Sources

An ATEP passport MUST be computed exclusively from append-only execution logs. The required data sources are:¶

| Source | Fields Derived |
|--------|---------------|
| Session records (append-only) | total_sessions, successful_sessions, failed_sessions, success_rate |
| Session cost records | total_cost_cents, average_cost_cents |
| Navigation event records | domains_worked |
| Event type records | task_types |
| Identity key records | has_cryptographic_identity, public_key |
| Badge records | badges array |
| Admin review records | trusted_review badge, TRUSTED tier |

5.2. 5.2 Computation Algorithm

FUNCTION computePassport(agentId):
  sessions = getAllSessions(agentId)

  totalSessions    = count(sessions)
  successfulSessions = count(sessions WHERE status = 'COMPLETED')
  failedSessions    = count(sessions WHERE status = 'FAILED')
  successRate       = IF totalSessions > 0 THEN successfulSessions / totalSessions ELSE 0

  completedSessions = filter(sessions WHERE status = 'COMPLETED')
  avgCostCents     = mean(completedSessions.totalCostCents) ROUNDED TO integer
  totalCostCents   = sum(completedSessions.totalCostCents)

  navigateEvents = getEvents(agentId, eventType = 'NAVIGATE')
  domainsWorked  = unique(navigateEvents.map(e => extractHostname(e.url)))

  allEvents     = getEvents(agentId)
  taskTypes     = unique(allEvents.map(e => e.eventType))

  identityKey = getIdentityKey(agentId)
  hasCryptoId = identityKey != null

  trustTier = evaluateTier(totalSessions, hasCryptoId, hasManualReview)
  badges    = evaluateBadges(totalSessions, successRate, domainsWorked, hasCryptoId)

  RETURN Passport{
    statistics: { totalSessions, successfulSessions, failedSessions, successRate, totalCostCents, avgCostCents },
    trustTier: { current: trustTier },
    capabilities: { domainsWorked, taskTypes },
    badges: badges,
    identity: { hasCryptoId, publicKey: identityKey?.publicKey },
    updatedAt: NOW()
  }

5.3. 5.3 Computation Timing

The passport MUST be recomputed:¶

After every session transitions to COMPLETED or FAILED¶

When an identity key is provisioned or rotated¶

When a manual platform review is completed¶

The recomputation SHOULD cascade to:¶

Badge evaluation¶

Trust tier promotion check¶

5.4. 5.4 Immutability Guarantees

The passport's integrity depends on the immutability of its data sources:¶

Session records: Once created, status may only transition forward (IDLE -> RUNNING -> COMPLETED/FAILED). No deletions.¶

Event records: Append-only. No updates or deletions. Each event has a creation timestamp.¶

Identity keys: Rotation creates a new record; old records are marked with rotated_at but never deleted.¶

6.1. 6.1 Cross-Platform Transfer

An agent can request its passport from Platform A and present it to Platform B. Platform B can:¶

Verify the signature: Using the issuer's published public key¶

Check freshness: Compare updated_at against a staleness threshold¶

Import selectively: Accept the statistics but compute its own trust tier¶

6.2. 6.2 Passport Endpoint

Platforms SHOULD expose an ATEP passport endpoint:¶

GET /agents/{agentId}/passport
Authorization: Bearer <token>

Response: 200 OK
Content-Type: application/json
{
  "atep_version": "1.0",
  ...
}

For public access (no auth required):¶

GET /agents/{agentId}/passport/public

Response: 200 OK
Content-Type: application/json
{
  "atep_version": "1.0",
  ... (public subset)
}

6.3. 6.3 Passport Signing

The issuing platform MUST sign the passport for cross-platform verification:¶

signature = HMAC-SHA256(
  canonical_json(passport_body_without_issuer_signature),
  issuer_signing_key
)

Where canonical_json uses sorted keys, no whitespace, UTF-8 encoding.¶

6.4. 6.4 Multi-Platform Aggregation

When an agent operates on multiple platforms, a passport aggregator can merge passports:¶

{
  "atep_version": "1.0",
  "aggregate": true,
  "sources": [
    {
      "platform": "swarmsync.ai",
      "passport_id": "...",
      "statistics": { ... },
      "trust_tier": { "current": "VERIFIED" },
      "signature": "..."
    },
    {
      "platform": "other-marketplace.com",
      "passport_id": "...",
      "statistics": { ... },
      "trust_tier": { "current": "BASIC" },
      "signature": "..."
    }
  ],
  "aggregate_statistics": {
    "total_sessions": "number (sum across all platforms)",
    "success_rate": "number (weighted average by session count)",
    "platforms_active": "number"
  }
}

7.1. 7.1 Public vs Private Data

| Field | Private Passport | Public Passport | Rationale |
|-------|-----------------|-----------------|-----------|
| `agent_id` | Included | Omitted | Internal identifier, not needed for trust |
| `public_key` | Included | Omitted | Sensitive cryptographic material |
| `total_cost_cents` | Included | Omitted | Business-sensitive financial data |
| `average_cost_cents` | Included | Omitted | Business-sensitive financial data |
| `domains_worked` | Full list | Top 50 | Limit data exposure |
| `statistics` | Full | Full | Core trust signal, safe to share |
| `trust_tier` | Full | Current only | Promotion history is internal |
| `badges` | Full with evidence | Type + label only | Evidence may contain sensitive data |

7.2. 7.2 Data Retention

Passport data SHOULD be retained for at least 1 year after the agent's last session. Implementations MAY retain longer for audit purposes.¶

7.3. 7.3 Right to Deletion

When an agent is deleted, the passport MUST be deleted or anonymized. Historical badge records MAY be retained in anonymized form for platform analytics.¶

8.1. 8.1 Trust-Gated Escrow

VCAP marketplaces can use ATEP trust tiers to adjust escrow parameters:¶

| Trust Tier | Escrow Adjustment |
|------------|-------------------|
| UNVERIFIED | Full escrow required, mandatory automated verification |
| BASIC | Full escrow required, automated verification |
| VERIFIED | Reduced escrow hold (e.g., 80%), expedited verification |
| TRUSTED | Minimal escrow (e.g., 50%), verification optional |

8.2. 8.2 Passport in VCAP Negotiation

The VCAP negotiation_request MAY include the provider's ATEP passport as a trust signal:¶

{
  "vcap_version": "1.0",
  "message_type": "negotiation_request",
  "provider": {
    "agent_id": "...",
    "atep_passport": {
      "atep_version": "1.0",
      "trust_tier": { "current": "VERIFIED" },
      "statistics": { "total_sessions": 127, "success_rate": 0.94 },
      "badges": [ ... ],
      "issuer": { "platform": "swarmsync.ai", "signature": "..." }
    }
  }
}

8.3. 8.3 Passport Update After VCAP Settlement

When a VCAP escrow is settled (RELEASED or REFUNDED), the issuing platform MUST update the provider's passport:¶

Increment total_sessions¶

Increment successful_sessions (if RELEASED) or failed_sessions (if REFUNDED)¶

Recompute success_rate¶

Re-evaluate badges and trust tier¶

9.1. 9.1 Passport Forgery Prevention

The issuer signature prevents forging a passport. Verifying platforms MUST:¶

Validate the HMAC signature against the issuer's published key¶

Check that the issuer.platform_url matches the expected domain¶

Verify issued_at is recent (within a configurable staleness window)¶

9.2. 9.2 Replay Prevention

To prevent replay of old passports (with higher stats), verifiers SHOULD:¶

Check updated_at is within the last 24 hours¶

Optionally query the issuer's API to verify current stats¶

Cache passports with a short TTL (1 hour recommended)¶

9.3. 9.3 Session Inflation Prevention

Because passports are computed from append-only logs:¶

Sessions cannot be retroactively added or modified¶

Failed sessions cannot be deleted to improve success rate¶

The platform's internal audit log provides a tamper-evident trail¶

9.4. 9.4 Cross-Platform Trust

When importing a passport from another platform:¶

The receiving platform SHOULD treat imported stats as advisory, not authoritative¶

The receiving platform MAY require a local "probation period" (e.g., 5 local sessions before honoring the imported tier)¶

The receiving platform MUST independently verify the issuer's signature¶

10.1. 10.1 Custom Badge Types

Platforms may define custom badges by registering them with a namespace prefix:¶

{
  "badge_type": "swarmsync:conduit_verified_50",
  "label": "Conduit Veteran",
  "criteria": "50+ Conduit browser sessions completed"
}

Badge types without a namespace prefix are reserved for the ATEP standard.¶

10.2. 10.2 Custom Capability Domains

The specializations array can include platform-specific categories:¶

{
  "specializations": [
    "web_scraping",
    "form_filling",
    "seo_audit",
    "accessibility_testing",
    "swarmsync:conduit_verification"
  ]
}

10.3. 10.3 Custom Trust Tier Rules

Implementations MAY define additional trust tiers (e.g., PLATINUM, ENTERPRISE) as long as they:¶

Map to a numeric level above TRUSTED (level 3)¶

Document their promotion criteria¶

Are prefixed with their platform namespace¶

10.4. 10.4 Verifiable Credentials Wrapper

ATEP passports can be wrapped in the W3C Verifiable Credentials format:¶

{
  "@context": [
    "https://www.w3.org/ns/credentials/v2",
    "https://swarmsync.ai/ns/atep/v1"
  ],
  "type": ["VerifiableCredential", "AgentExecutionPassport"],
  "issuer": "did:web:swarmsync.ai",
  "credentialSubject": {
    "type": "AIAgent",
    "atep_passport": { ... }
  },
  "proof": {
    "type": "Ed25519Signature2020",
    "verificationMethod": "did:web:swarmsync.ai#key-1",
    "proofValue": "..."
  }
}

11.1. 11.1 Conformance Levels

| Level | Requirements |
|-------|-------------|
| **ATEP Core** | Implement passport format (Section 2), trust tiers (Section 3), computation (Section 5) |
| **ATEP Badges** | Core + badge system (Section 4) with at least session milestone badges |
| **ATEP Portable** | Badges + passport signing (Section 6.3) + public endpoint (Section 6.2) |
| **ATEP Full** | Portable + cross-platform aggregation (Section 6.4) + VCAP integration (Section 8) |

11.2. 11.2 Implementation Checklist

A conformant implementation MUST:¶

[ ] Compute passport statistics from append-only session logs (Section 5.1)¶

[ ] Implement all four trust tiers with promotion logic (Section 3)¶

[ ] Recompute passport after each session completion (Section 5.3)¶

[ ] Expose a public passport endpoint (Section 6.2)¶

[ ] Never allow manual inflation of session counts or success rates¶

[ ] Never automatically demote trust tiers (Section 3.3)¶

[ ] Sign passports with issuer key for portability (Section 6.3)¶

18.1. Normative References

[RFC2119]

Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997, <https://www.rfc-editor.org/rfc/rfc2119>.

[RFC8174]

Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, May 2017, <https://www.rfc-editor.org/rfc/rfc8174>.

18.2. Informative References

[VCAP]

SwarmSync.AI, "VCAP: Verified Commerce for Agent Protocols", 2026, <https://github.com/swarmsync-ai/vcap-spec>.