Trust Levels & Badges

Understanding Cargo trust levels and what badges mean

Trust Levels & Badges

Every Cargo on Archipelag.io has a trust level (0-3) displayed as a badge. This page explains what each level means and how to interpret trust signals when choosing Cargos.

Trust Level Overview

Level 0: Sandbox

Badge: Gray shield

What it means:

• Cargo was recently submitted
• May not have completed security scanning
• Runs with maximum restrictions

Capabilities:

• No network access
• Minimal resources (256MB RAM, 1 CPU)
• 60-second maximum duration
• Cannot access GPU

When you’ll see this:

• Brand new Cargos awaiting review
• Cargos that failed security requirements
• Cargos from Individual-tier publishers

Level 1: Community

Badge: Blue shield with checkmark

What it means:

• Passed automated security scanning
• No critical or high-severity vulnerabilities
• Has been used successfully by the community (100+ jobs)
• Publisher is at least Verified tier

Capabilities:

• Standard sandbox (1GB RAM, 2 CPU)
• 5-minute maximum duration
• No network access
• No GPU access

Trust signals to look for:

• Job success rate (aim for >95%)
• Number of jobs completed
• Publisher verification status
• User ratings and reviews

Level 2: Verified

Badge: Green shield with checkmark

What it means:

• Passed automated security scanning
• Manually reviewed by platform staff
• Cryptographically signed by Archipelag.io
• Build provenance verified (SLSA attestation)
• Publisher is Organization tier or higher

Capabilities:

• Elevated sandbox (8GB RAM, 4 CPU)
• 30-minute maximum duration
• Outbound network access (for API calls, etc.)
• No GPU access (unless Partner tier)

How Cargos reach this level:

1. Publisher requests staff review
2. Platform team reviews code, container, and behavior
3. If approved, Cargo is signed with platform key
4. Signature is verified by Islands before execution

Level 3: Official

Badge: Purple star

What it means:

• Built and maintained by Archipelag.io
• Full platform backing and support
• Highest level of verification
• Used for core platform capabilities

Capabilities:

• Custom resource limits
• GPU access
• Network access
• Priority scheduling

Examples:

• archipelag/llm-chat - LLM inference
• archipelag/image-gen - Image generation
• archipelag/embeddings - Vector embeddings

Reading Trust Information

When viewing a Cargo, click the trust badge to see detailed information:

Trust Details Modal

The trust modal shows:

Verification Status

• Current trust level and how it was achieved
• Security scan results (vulnerability counts by severity)
• Signature status (signed/unsigned, key fingerprint)
• Build provenance (if available)

Publisher Information

• Publisher name and tier
• Verification method (Stripe, GitHub, Domain)
• Time on platform
• Other published Cargos

Performance Metrics

• Reputation score (0.0 - 5.0)
• Total jobs completed
• Success rate percentage
• Average response time

Security Scan Summary

• Scan date
• Scanner tool and version
• Vulnerability breakdown:
  • Critical: Must be 0 for Level 1+
  • High: Must be ≤5 for Level 1, 0 for Level 2+
  • Medium: Informational
  • Low: Informational

Evaluating Cargos

For Personal Use

Community Cargos (Level 1) are generally fine for personal projects and experimentation. Check:

• Publisher has Verified tier or higher
• Success rate is above 90%
• No recent complaints or issues

For Production Use

For production applications, prefer Verified Cargos (Level 2):

• Cryptographic signature ensures integrity
• Staff review adds human oversight
• Build provenance enables supply chain verification

Red Flags to Watch For

Be cautious of Cargos with:

• Sandbox level (Level 0) with no path to verification
• Success rate below 85%
• Reputation score below 1.0
• Many recent failures or complaints
• Publisher with no other Cargos
• Vague or missing descriptions

How Trust Levels Change

Promotion

Cargos can be promoted to higher trust levels:

Demotion

Cargos can be demoted for:

• New security vulnerabilities discovered in rescans
• Success rate drops below threshold
• Reputation score falls below minimum
• Policy violations
• Publisher suspension

Sandbox Tiers (Technical)

Trust levels map to sandbox tiers that control runtime restrictions:

These restrictions are enforced at the Island level using container isolation, seccomp profiles, and resource cgroups.

Frequently Asked Questions

Can I request a trust level upgrade?

For Level 1 (Community), upgrades happen automatically when requirements are met. For Level 2 (Verified), Organization-tier publishers can request staff review from the Cargo settings page.

How often are Cargos rescanned?

All Cargos are rescanned weekly. If new vulnerabilities are found, trust levels may be adjusted automatically.

What if I disagree with a trust level decision?

Contact support@archipelag.io with your Cargo ID and concerns. We’ll review the decision.

Do trust levels affect pricing?

No, pricing is based on resource usage, not trust level. However, higher trust levels unlock more resources, which may affect total cost for resource-intensive jobs.

Next Steps

{% end %}