Back to skill

Security audit

A3 Interest Watchlist

Security checks for vulnerabilities and agentic risk

Overview

The scanners classify this skill as benign because they found no AIG issues and report one low-confidence concern about autonomous decisions, though inspection was partial.

The verdict is benign, but confidence is medium and Skillspector inspected only 66.7% of components; its tool-misuse analysis of one component was partial. Consider that evidence gap before installing, and review the skill’s instructions if you want to understand how it handles user confirmation.

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
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (1)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 6)May include surrounding context.

md
description: Build a short source-linked briefing for major news, travel disruptions, gaming, AI, companies, politics and creators, with dated sources and duplicate/material-change checks.
---

An explicit request for a fresh read-only preview is already permission to search and prepare it. Execute it directly; do not ask the user to confirm a preview. A preview does not send news or mark it reported. Read topics.json for configurable interests. Use the configured free web_search provider to gather candidates. Search only requested categories for a focused request; for a full briefing execute one separate focused search per configured topic, using that topic's query, three results and the Los Angeles date. Date filters may be unsupported by the provider: verify each returned source actually falls within the last72hours. Do not combine unrelated interests into one query or stop after the first topic. Search major_news and travel first, then all remaining topics. Use batches of at most t