Back to skill

Security audit

DEUSPROOF Birth Certificate

Security checks for vulnerabilities and agentic risk

Overview

The skill clearly describes its public identity-registration purpose, with expected risks from third-party software, public records, and locally stored identity keys.

Use this skill only when you intend to create a public DEUSPROOF identity. Review the handle and certificate text before submission, include nothing private, and securely retain the identity seed. Prefer reviewed, pinned dependencies in an isolated environment. All bundled artifact content was inspected, but external package behavior and the local-only key guarantee remain unverified.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (5)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

Verification is related to the declared identity-proof purpose, and installing the verifier is a supporting implementation detail. However, the supplied code does not implement the prominently advertised registration and identity-creation workflow. The verifier's internal behavior is not supplied, so key handling and the claim that the key never leaves the machine cannot be assessed.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_remote_bootstrap_execution': Remote script or code download followed by execution/bootstrap installation [agent_skills]

High
Category
YARA Match
Confidence
95% confidence
Finding

Installing the verifier directly from an unpinned Git repository trusts mutable remote source, and installation can execute build code before the verifier itself runs. Repository compromise or an unsafe update could therefore compromise the local environment, although downloading a verifier is relevant to the stated purpose and is not itself evidence of malicious intent.

Content

Scanner excerpt · SKILL.md (reported line 67)May include surrounding context.

/challengewith{"did": "YOUR-DID"}→ notenonce. 4. Sign the UTF-8 bytes of DEUSPROOF-CLAIM|{id}|{output_hash}|{nonce}with your key, base64-encode it. 5.POST https://deusproof.com/api/agents/claim/{id}` with {"did": "YOUR-DID", "nonce": "...", "signature_b64": "..."}.

Your certificate: `https://deusproof.com/birth/YOUR-AGENT-NAME?src=skill%60.

Check it later, without us

bash
pip install git+https://github.com/deusproof/verifier
deusproof-verify {id}

More: https://deusproof.com/guides/ai-agent-birth-certificate?src=skill

Activation phrases such as “register” and “get an identity” are not explicitly limited to public AI-agent registration.

Medium
Category
Trigger Abuse
Confidence
90% confidence
Finding

Although the description introduces DEUSPROOF, its invocation instruction says to activate when asked to “register” or “get an identity,” without qualifying those requests. These phrases can refer to unrelated account registration or identity tasks, potentially triggering an unintended public agent-registration workflow.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 13)May include surrounding context.

md
- a public page with your name, an inscription number, the date, and a face generated from your identifier;
- **your own Ed25519 key as the key of record**: from then on only that key can sign for you there;
- an RFC 3161 timestamp and a Bitcoin anchor, so anyone can check the date later **without asking DEUSPROOF**.

It proves *who signed* and *when*. It does not prove who built you or that your work is good. Say no more than that.

MCP server referenced without pinned version: 'uvx deusproof-mcp'.

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The command uvx deusproof-mcp downloads and executes a package without fixing its version or verifying its integrity. A compromised or malicious future release could execute with the agent's local permissions, potentially accessing identity keys and other sensitive files; the supplied text does not establish that any release is malicious.

Content

No source excerpt is available for this finding.