Security audit

七宗罪.skill / 7 Sins.skill

Security checks for vulnerabilities and agentic risk

Overview

This is a prompt-only critique skill whose web-search use and harsh evaluative style are disclosed and aligned with its stated purpose.

Install this only if you want an intentionally blunt critique style. Avoid submitting confidential product plans if your host environment enables web search, because parts of the analysis may use public internet lookups.

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

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

Findings (4)

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The skill enables implicit invocation but provides no narrowly defined trigger conditions or scope limits. That increases the chance the agent will invoke this adversarial, high-pressure 'judgment' skill in unrelated contexts, causing unwanted steering of user decisions or inappropriate responses without clear user intent.

Natural-Language Policy Violations

Low

Confidence: 93% confidence
Finding: The instruction "Keep runtime instructions English-first" establishes a preferred language policy that is not framed as user opt-in. Although later lines say to localize visible answers to the user's dominant language, this line still imposes a specific language constraint that can conflict with a user's locale preference.

Missing User Warnings

Low

Confidence: 88% confidence
Finding: The skill instructs Greed and Envy to perform web searches and use external evidence, but the description does not clearly warn users that external lookups may occur. This can undermine user expectations around privacy, confidentiality, and reproducibility, especially if users submit sensitive product plans, code, or business ideas assuming analysis is fully local.

Natural-Language Policy Violations

Low

Confidence: 83% confidence
Finding: The manifest presents user-facing name and descriptions bilingually with mandatory Chinese content, but does not indicate that language choice is optional or region-specific. This may violate language/locale policy expectations where users should be offered a choice rather than having a locale imposed implicitly.

Static analysis

No suspicious patterns detected.