Security audit

Leo's Reading Notes

Security checks for vulnerabilities and agentic risk

Overview

This reading-notes skill is mostly coherent, but it can automatically save personal and work-related reflections, use external search, and set up ongoing reminder behavior from broad triggers without clear user control.

Review this skill before installing if you discuss books casually or do not want personal reflections, work context, future plans, and action items saved by default. Use it only if you are comfortable with local note persistence, external searches for book text, and a daily internalization workflow, and consider requiring explicit confirmation before any search or file write.

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

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (7)

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The trigger conditions are broad enough to activate on ordinary discussion about books or ideas, causing the agent to enter a mode that performs storage and other side effects without an explicit opt-in at that moment. This increases the risk of unintended collection, persistence, and workflow changes from casual conversation.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill instructs external searching for original book text based on user input without warning that user-provided content may be sent to external services. This can leak reading interests, quoted text, personal interpretations, or sensitive context to third parties without transparency or consent.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The workflow directs the agent to create or update files automatically and then merely inform the user afterward, which removes informed consent for filesystem changes. An attacker or accidental trigger could therefore cause silent modification of persistent data, including overwriting or proliferating files.

Ssd 3

Medium

Confidence: 94% confidence
Finding: The skill requires collecting and storing users' personal reflections, work context, future plans, and action items as part of a mandatory note structure. This creates a persistent repository of potentially sensitive personal and professional information without clear data minimization, retention limits, or consent boundaries.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The skill directs the agent to append detailed user responses into persistent note files and maintain a progress tracker of internalization status. This expands data collection from simple notes into behavioral tracking over time, increasing privacy risk and the consequences of unauthorized access or misuse.

Ssd 3

Medium

Confidence: 94% confidence
Finding: The example explicitly instructs writing the user's internalization content back into files, reinforcing a default behavior of persisting introspective and work-related material. Because the content can include goals, habits, and business intentions, the storage behavior raises confidentiality and profiling concerns beyond ordinary note-taking.

Natural-Language Policy Violations

Low

Confidence: 95% confidence
Finding: This shell script includes user-facing comments and status output entirely in Chinese, such as the publish progress and success/failure messages. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation when no locale choice or justification is provided.

Static analysis

No suspicious patterns detected.