Security audit

deco-biz-assistant

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed renovation-business drafting skill with local data and human-confirmation boundaries, and I found no hidden code or unsafe install behavior.

Before installing, confirm that your agent's project archive, price library, and writable memory are limited to the intended renovation-business workspace. Treat generated quotes, bids, and proposals as drafts, and keep the stated human approval step for price updates, final documents, and anything sent outside the company.

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

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The skill’s routing step says to classify any received instruction into one of several modules, but it does not define strong entry conditions, allowed verbs, or rejection criteria. That can cause the skill to activate on broad business-language prompts and process unrelated or sensitive requests under an overly permissive scope, increasing the chance of unintended data access or generation of business documents from ambiguous input.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger list contains broad, high-frequency business terms such as “报价”, “方案书”, and “招标”, which can cause the skill to activate in contexts where the user did not intend to invoke this workflow. In an agent with document generation, project archive access, and writable memory, unintended invocation can expose internal data, create misleading drafts, or initiate state changes without clear user intent.

Static analysis

No suspicious patterns detected.