Security audit

Paper Engineering Assistant

Security checks for vulnerabilities and agentic risk

Overview

This academic writing skill uses disclosed project-file writes and literature-search networking that fit its stated purpose, with some care needed around overwrites.

Install only if you want an assistant that can create and update files in an academic writing project. Set an explicit project directory, keep backups of thesis drafts, and ask for a preview before operations that may overwrite body files or download literature.

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 (3)

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The trigger examples are broad enough that normal academic-assistance requests like 'start working on my thesis' could activate a workflow with filesystem and network side effects. Because the skill can read/write project files, overwrite body content, and download literature, ambiguous activation increases the chance of unintended execution and excessive access beyond the user's precise intent.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The invocation logic accepts underspecified requests such as 'start,' 'process references,' or 'find literature' and then performs multi-step actions over directories, existing files, and network resources. In this skill, those vague requests can lead to broad file enumeration, automatic synchronization, overwriting of existing content, and external downloads without strong scoping or per-action consent, making accidental destructive or privacy-impacting behavior more likely.

Missing User Warnings

Low

Confidence: 88% confidence
Finding: This code overwrites the summary file on disk using json.dump, but the operation has no confirmation prompt and no docstring or comment warning that a persistent file write will occur. The returned status message comes after the write and does not meaningfully disclose the safety-relevant behavior beforehand.

Static analysis

No suspicious patterns detected.