Security audit

独立思考五部曲

Security checks for vulnerabilities and agentic risk

Overview

This skill is not clearly malicious, but it gives an agent broad private-memory lookup and mandatory wiki persistence without clear user control.

Install only if you trust the referenced local vector script and the wiki/vector stores it will access. Prefer editing the skill so vector searches and wiki writes require explicit user approval, remove the permanent Memory Search override, and document storage location, retention, deletion, and review controls.

SkillSpector

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
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The skill explicitly requires writing outputs back into a wiki, which changes persistent state rather than merely guiding internal reasoning. That creates a data-governance and integrity risk: user-derived or model-generated content may be stored long-term without validation, consent, or clear scope controls.

Vague Triggers

Medium

Confidence: 79% confidence
Finding: The trigger conditions are broad enough to cover many normal knowledge-work tasks such as troubleshooting, decisions, retrospectives, and uncertain topics. In practice this can cause the skill to activate frequently, expanding when vector lookup and wiki interaction occur and increasing exposure of user content to retrieval and persistence flows.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill mandates use of a specific vector retrieval command and wiki access before responding, but provides no disclosure that user prompts or derived queries may be sent to another knowledge system. This creates privacy and policy risks because sensitive user data may be propagated to local or external stores without transparency or minimization.

Missing User Warnings

High

Confidence: 98% confidence
Finding: The requirement that outputs 'must' be stored into the wiki creates mandatory persistence of user-influenced content without warning, consent, or review. This is more dangerous than transient retrieval because it can permanently retain sensitive, incorrect, or harmful content and feed it back into future decisions.

Static analysis

No suspicious patterns detected.