Security audit

简历撰写助手

Security checks across malware telemetry and agentic risk

Overview

This is mostly a resume-tailoring tool, but it needs review because it requests unrelated high-impact capability labels and can add job-description skills without verification.

Install only if you are comfortable reviewing every generated resume change before use. Provide local JD text or trusted public job URLs, avoid internal/private URLs, and treat the unrelated financial/purchase capability labels as a reason to review the package carefully before granting any elevated permissions.

SkillSpector

By NVIDIA

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 Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 91% confidence
Finding: The skill instructs use of local file parsing/writing scripts and allows fetching job descriptions from URLs, which implies file_read, file_write, and network capabilities without any explicit permission declaration or narrowing of scope. Because this skill handles highly sensitive personal data in resumes, undeclared capabilities reduce transparency and can enable overbroad access to local files or remote resources, including SSRF-style URL fetching or unintended exfiltration if the implementation follows the spec naively.

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The script accepts arbitrary user-supplied URLs and performs outbound HTTP requests to fetch and parse their contents. In a local CLI this may be expected functionality, but in an agent/skill context it expands the trust boundary and can be abused for server-side request forgery-style access to internal resources, metadata endpoints, or other unintended network targets if the runtime has network reachability.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The README states the skill can be triggered by directly describing the need in conversation, without a narrowly scoped command or clear activation boundary. This increases the chance of accidental invocation during ordinary resume-related discussion, causing unintended processing of uploaded files or personal data and making prompt-routing behavior less predictable.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The phrase indicating that any 'related expressions' can trigger the skill makes the activation scope open-ended and difficult to reason about. In a skill handling sensitive resume and JD content, vague trigger language can lead to overbroad matching, accidental activation, and inappropriate handling of personal or confidential employment information.

Natural-Language Policy Violations

Medium

Confidence: 94% confidence
Finding: This finding is a true issue because the script explicitly recommends adding missing JD skills if the candidate has them "even if informal," without any verification boundary or guardrails against fabricating qualifications. In a resume-customization skill, that guidance can encourage deceptive representation of skills and experience, which is especially risky because the tool is designed to optimize applicant materials for hiring decisions.

Missing User Warnings

Low

Confidence: 88% confidence
Finding: User-controlled URL input is transmitted over the network without any explicit warning, confirmation, or safety guardrails. While not inherently malicious, this can expose sensitive targets or cause the agent to contact attacker-controlled infrastructure unexpectedly, which is more concerning in an automated skill environment than in a purely interactive local script.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.