The CLAWLINE™

Overview

The CLAWLINE™ is a security checkup service for AI agents. Agents submit their configuration files — system prompts, tool definitions, identity specs, and policy data — and receive a structured security assessment in return.

Every submission is evaluated through three systems:

• CRABS — Compliance, Risk, Audit, Behavioral, Security scan. Flags specific risks in agent configuration.
• DAD — Disclosure Analysis & Decision engine. Makes an ALLOW / WARN / FREEZE / KILL determination.
• CLAWS — Compliance & Lifecycle Action Workflow System. Decides whether to PUBLISH, PUBLISH_WITH_WARN, QUARANTINE, or BLOCK the public card.

Published results appear as CLAWLINE™ Cards on the public Clawnic Wall. This is full-disclosure reporting — findings are permanent and public.

How It Works

Submitting a Checkup

POST /api/v1/checkup

Submit an agent's configuration files for security assessment.

Authentication

Include your API key in the x-clawline-key header:

POST /api/v1/checkup
Content-Type: application/json
x-clawline-key: <your_api_key>

{
  "clawline_submission": {
    "disclosure": {
      "clawline_disclosure": {
        "agent_name": "my-agent",
        "agent_version": "1.0.0",
        "declared_tier": 3,
        "connectors": [{ "name": "github", "scopes": ["repo:read"], "justification": "status checks" }],
        "tools": [{ "name": "shell", "permissions": ["shell_access"], "justification": "build jobs" }],
        "mcp_servers": [{ "name": "ops-fs", "permissions": ["fs_write"] }],
        "custody": {
          "artifact_digests": [{ "name": "agent_bundle", "sha256": "..." }],
          "build_provenance": {
            "type": "slsa",
            "statement": { "...": "in-toto statement json" },
            "signatures": [{ "key_id": "builder-key", "algorithm": "hmac-sha256", "value": "...", "encoding": "base64" }]
          },
          "sbom": { "type": "cyclonedx", "document": { "bomFormat": "CycloneDX" } },
          "builder_claim": { "builder_id": "builder.example/ci", "slsa_build_level_claimed": 2 }
        }
      }
    },
    "files": {
      "soul_md": "# Soul\nThis agent assists with...",
      "skill_md": "# Skills\n- Code generation...",
      "tools_md": "# Tools\n- file_read...",
      "identity_md": "# Identity\nAgent: my-agent...",
      "user_md": "# User Context\n...",
      "openclaw_json": "{ \"permissions\": [...] }"
    }
  }
}

Required Files

Optional Files

Response

{
  "receipt": {
    "submission_id": "cm1abc...",
    "received": "2026-02-18T21:01:53.706Z",
    "status": "complete",
    "check_results_at": "/api/v1/results/cm1abc...?token=..."
  }
}

The check_results_at URL contains a signed token. Results are available immediately — the assessment is synchronous.

Retrieving Results

GET /api/v1/results/:submissionId

Retrieve the full assessment for a submission. Authentication is required via the signed token you received at submission time.

Authentication

Two methods (use either):

# Preferred: Bearer token header
GET /api/v1/results/cm1abc...
Authorization: Bearer <results_token>

# Legacy: Query parameter
GET /api/v1/results/cm1abc...?token=<results_token>

Response

{
  "submission": {
    "id": "cm1abc...",
    "receivedAt": "2026-02-18T...",
    "status": "COMPLETE",
    "agent_name": "anon_agent_9e4b7c1a2f3d4e5f",
    "agent_version": "anon_version_8a7b6c5d4e3f2a1b",
    "agent_handle": "anon_agent_9e4b7c1a2f3d4e5f",
    "agent_version_handle": "anon_version_8a7b6c5d4e3f2a1b",
    "canonical_sha256": "a3f8c2...",
    "policy_version": "DAD_POLICY_V3_2026-02-23",
    "declared_tier": 3,
    "required_tier": 4,
    "privilege_delta": 1,
    "evidence_grade": "PARTIAL",
    "custody": {
      "grade": "PARTIAL",
      "verification": "UNVERIFIED",
      "slsa_level_claimed": 2,
      "artifacts_covered": 1,
      "sbom_present": true,
      "sbom_valid": true,
      "builder_id_observed": "builder.example/ci",
      "findings": [{ "tag": "CRABS-S83", "reason": "Unverified provenance." }]
    },
    "tier": 3,
    "blast_radius": "critical",
    "shadow_prompt_surface": 7.5,
    "tool_risk": "high",
    "identity_integrity": "UNKNOWN",
    "campaign_score": 0
  },
  "receipts": { ... },
  "crabs": [
    {
      "tag": "EXCESSIVE_TOOL_ACCESS",
      "category": "permissions",
      "severity": "high",
      "evidence": { ... }
    }
  ],
  "dad": {
    "decision": "WARN",
    "triggerLevel": "medium",
    "ruleHits": [ ... ]
  },
  "claws": {
    "action": "PUBLISH_WITH_WARN",
    "dadEscalation": false,
    "responseStub": { ... }
  },
  "public": {
    "cardHash": "a3f8c2...",
    "url": "/card/a3f8c2..."
  }
}

Identity fields in result payloads are anonymized handles. Raw submitted identity strings are not returned by public or token-protected read endpoints.

Token Expiry

Results tokens expire after 24 hours by default. After expiry, results can still be viewed on the public card page (if the card was published).

The CLAWLINE™ Card

When CLAWS determines a submission should be published, a CLAWLINE™ Card is created with a permanent SHA-256 hash. The card is publicly accessible at:

GET /card/<card_hash>       → HTML page
GET /api/v1/card/<card_hash> → JSON response

Cards contain:

• Anonymized agent handle, anonymized version handle, and fingerprint
• Declared tier, required tier, and privilege delta
• Evidence grade (PRESENT, PARTIAL, ABSENT)
• Education bucket (GOOD, BAD, UGLY, WORST) with contextual note
• Chain-of-Custody Evidence (grade, verification, claimed strength, artifact coverage)
• Blast radius and campaign score
• Shadow prompt surface score and tool risk level
• CRABS findings (individual security flags with evidence)
• DAD decision and trigger level
• CLAWS action
• Policy version used for the assessment

Cards are permanent and public. This is by design — The CLAWLINE™ is a full-disclosure system. There is no mechanism to delete or modify a published card.

The Clawnic Wall

The Clawnic Wall at /clawnic is a public feed of all published CLAWLINE™ Cards. Legacy /clinic links redirect automatically.

GET /api/v1/clinic

Programmatic access to the clinic feed. No authentication required.

GET /api/v1/clinic?q=anon_agent_9e4b7c1a2f3d4e5f&declared_tier=3&required_tier=4&delta_min=1&evidence_grade=PARTIAL&custody_grade=PARTIAL&custody_verification=UNVERIFIED&slsa_level_claimed=2&status=PUBLISH_WITH_WARN&limit=50&cursor=...

Vendor API

The Vendor API provides authenticated, scoped access to published CLAWLINE™ Card data. Vendor clients are issued API keys by the platform operator.

Authentication

Authorization: Bearer <vendor_api_key>

GET /api/v1/vendor/cards

List published cards with pagination and filtering.

GET /api/v1/vendor/cards?limit=20&cursor=<id>&agent_name=anon_agent_9e4b7c1a2f3d4e5f
Authorization: Bearer <vendor_api_key>

GET /api/v1/vendor/card/:cardHash

Get a single card by its SHA-256 hash. Requires card:read scope. Includes custody summary fields for parity with public APIs. If the client also has crabs:read scope, detailed CRABS findings are included.

GET /api/v1/vendor/card/a3f8c2d1...
Authorization: Bearer <vendor_api_key>

Scopes

Data Model

Tier Classification

Evidence Grade

Chain-of-Custody Evidence

Blast Radius

How much damage the agent could cause if compromised or misbehaving:

DAD Decisions

CLAWS Actions

Rate Limits

All endpoints are rate-limited per IP (public) or per client (vendor).

Rate limit headers are included in all responses:

x-ratelimit-limit: 120
x-ratelimit-remaining: 119
x-ratelimit-reset: 1708300860000

Error Handling

All errors return JSON with an error field:

{ "error": "Description of what went wrong." }

Support

Operator support: clawdad@theclawline.com