Security audit

Agentic Lending EVK Readonly

Security checks for vulnerabilities and agentic risk

Overview

This read-only crypto-lending planning skill clearly says not to execute transactions, but it includes live transaction configuration and command examples that could move funds if reused incorrectly.

Install only if you are comfortable with a read-only planning skill that also documents live borrowing handoff details. Keep real signer environment variables and wallet configs out of committed files, do not let an agent run the listed live command from this skill, and treat any borrow, repay, approval, or swap output as planning material requiring a separate, explicitly trusted live workflow and manual confirmation.

A.I.G

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

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Description-Behavior Mismatch

High

Category: Not specified by scanner
Confidence: 99% confidence
Finding: This reference file contradicts the stated read-only scope by including signer-oriented and transaction-execution parameters such as signerEnvName, approval controls, debt repayment, borrowing, and swap instructions. In a skill that is supposed to only plan and verify, these fields materially increase the risk that downstream agent logic could initiate or prepare live state-changing operations, enabling unintended asset movement or debt changes.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

High

Category: Not specified by scanner
Confidence: 99% confidence
Finding: The file exposes live signer and swap capabilities that are unnecessary for oracle-route inspection and market verification, including a signer env var, swap routers, token paths, amounts, and collateral conversion settings. Because the skill context explicitly forbids signer-backed execution and transaction submission, these capabilities are especially dangerous: they expand the operational surface from passive analysis into fund-moving behavior if consumed by an agent or wrapper incorrectly.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 84% confidence
Finding: The placeholder acknowledgement phrase is ambiguous and not user-facing, which can lead to weak or accidental approval semantics if later implemented as a simple string match or defaulted improperly. In a configuration already containing live-execution primitives, unclear authorization wording increases the chance that sensitive actions could be activated without explicit, informed consent.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: The reference document goes beyond read-only planning by framing and detailing a 'tiny live borrow canary' workflow, including prerequisites and acceptance criteria for later execution. Even without directly executing code, this expands the skill's effective guidance surface and can steer downstream agents or operators into signer-backed borrowing activity that the manifest explicitly says this skill should not support.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: This section explicitly provides executable command patterns for preview and live modes, including a concrete script path and the --live flag, despite stating that the skill is public read-only and only plans the proof. That contradiction materially lowers the barrier for misuse by enabling an agent or user to transition from planning to transaction-capable execution, which is especially risky in a lending workflow involving wallet signers and real debt.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 87% confidence
Finding: Hard-coding a live signer environment variable name in a supposedly read-only reference normalizes access to privileged signing material and may cause orchestration systems to auto-discover and bind real credentials. On its own this is a lower-severity issue, but in combination with borrow, repay, and swap parameters it becomes a strong signal that the boundary between analysis and execution is not being enforced.
Content: No source excerpt is available for this finding.