Security audit

Evomap Verify Report

Security checks for vulnerabilities and agentic risk

Overview

This is a small EvoMap report-submission skill with disclosed external posting behavior, but users should confirm the destination before sending report data.

Install only if you intend to submit EvoMap verification reports. Before sending, confirm the request is specifically for EvoMap and review the asset ID, pass/fail result, confidence, GDI score, comments, destination endpoint, and any stated USDC fee.

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
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The trigger phrase "submit report" is generic enough to activate this skill for unrelated user intents, which can cause unintended collection of asset IDs and verification data and route users into an external submission flow they did not explicitly request. In a skill that sends data to a third-party API, ambiguous invocation increases the risk of accidental data disclosure or unintended network actions.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill instructs collection of user-provided report fields and submission to an external endpoint without clearly warning that this information will leave the local agent context and be transmitted over the network. Users may disclose sensitive identifiers, comments, or assessment data without informed consent, especially because the workflow directly proceeds from prompt matching to payload submission.

Natural-Language Policy Violations

Low

Confidence: 80% confidence
Finding: The description and usage instructions are partially written in Chinese/Cantonese while also using English trigger phrases. This can impose a language expectation on users without explicit opt-in or a stated regional justification.

Static analysis

No suspicious patterns detected.