Security audit

Drawing Analyzer

Security checks for vulnerabilities and agentic risk

Overview

The skill appears intended for construction drawing analysis, but it asks for execution capability and uses an external PDF package while its activation and dependency disclosures are too broad and inconsistent.

Review before installing. This skill is not showing malicious behavior, but it should ideally narrow activation to construction drawing/PDF analysis, document pdfplumber consistently, and limit exec use to explicit user-approved install or analysis commands.

SkillSpector

Vulnerability Patterns

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
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (4)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The manifest makes strong security claims such as removing risky code and external dependencies, but the skill still declares executable capability and the implementation imports and relies on pdfplumber. This mismatch can cause users or platforms to trust the skill more than warranted, reducing scrutiny and increasing the chance that unnecessary execution or dependency risk is accepted.

Context-Inappropriate Capability

Medium

Confidence: 91% confidence
Finding: The skill declares the exec tool even though the included implementation only performs parsing and reporting logic. Unused execution capability expands the attack surface because an agent may permit shell or process execution for a skill that does not need it, enabling abuse if prompt instructions or future modifications invoke commands.

Intent-Code Divergence

Low

Confidence: 83% confidence
Finding: The dependency documentation states that no extra external setup is needed beyond the built-in LLM API, but the actual code requires pdfplumber. This is primarily a transparency and supply-chain hygiene issue: operators may deploy the skill without understanding its Python package dependency and associated patching, review, and installation requirements.

Vague Triggers

High

Confidence: 89% confidence
Finding: The trigger keywords are very broad terms like drawing, analyze, and extract, which are likely to match many ordinary user prompts unrelated to construction-document parsing. Over-broad activation can cause unintended skill invocation and, in this case, may expose unnecessary read/exec-capable behavior in contexts where the user did not intend to enable this skill.

Static analysis

No suspicious patterns detected.