Security audit

积极心理学工具箱

Security checks for vulnerabilities and agentic risk

Overview

This positive-psychology skill is coherent and local-only, but it can save sensitive mood and self-assessment data in plaintext files without clear opt-in or deletion controls.

Review before installing if you may enter private emotional or self-assessment information. Use it only on a trusted private machine, assume the Python tools create local plaintext files in the skill's scripts directory, delete those files when no longer needed, and do not rely on this skill for diagnosis, treatment, or crisis support.

SkillSpector

Vulnerability Patterns

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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The script stores a user's personality-assessment output to a local JSON file containing psychologically sensitive self-report data and timestamps, but it does so by default without an explicit consent flow, retention policy, or access controls. In the context of a mental-wellness skill, this increases privacy risk because users may not expect durable storage of intimate assessment results on shared machines, synced folders, backups, or agent-host environments.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The trigger scope is broad enough to activate on generic emotional-support, mood, self-growth, or life-guidance requests, which can cause this skill to engage outside a clearly bounded positive-psychology context. In a mental-health-adjacent skill, over-triggering is risky because users in distress may receive a framework-oriented wellness response when they instead need crisis handling, a different specialist skill, or stronger safety routing.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: This file provides mental-health exercises and even advises matching practices to the user's psychological state, but it does not include a clear crisis escalation warning for users with severe depression, self-harm risk, suicidality, or acute psychiatric distress. In this context, a user in crisis could be inappropriately guided into self-help exercises instead of being directed to urgent professional or emergency support, creating a meaningful safety risk.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The script persists highly sensitive mental-health-adjacent data, including emotions, triggers, bodily sensations, and context, to a predictable local JSONL file next to the script without access controls, encryption, retention limits, or an explicit consent/privacy prompt at the point of collection. In this skill context, the data is especially sensitive because it can reveal psychological state, personal relationships, routines, and distress patterns, so unauthorized local access or accidental sharing could cause meaningful privacy harm.

Missing User Warnings

Low

Confidence: 98% confidence
Finding: The script writes assessment results to disk automatically and only informs the user afterward, preventing informed consent. While this is primarily a privacy/UX issue rather than code execution risk, it is meaningful here because the saved content reflects mental-health-related personal profiling data.

Static analysis

No suspicious patterns detected.