article-add

================================================================================

🔵 VERDICT: LOW

This skill, article-add, is described entirely in Markdown and does not contain any executable scripts or direct commands. Its purpose is to store user-provided topics into a markdown file (article-queue.md) located in either ~/.claude/ or .claude/ directories. The skill itself does not perform any network operations, access sensitive files, or attempt privilege escalation.

The primary security consideration is the potential for Indirect Prompt Injection. The skill takes user input (<topic>) and writes it directly into a file. If a subsequent LLM-driven skill (e.g., the /article skill mentioned) reads and processes this file without adequate sanitization, a malicious topic could contain instructions designed to manipulate the LLM's behavior. This is a risk inherent in systems where LLMs process user-controlled data from persistent storage. However, this risk is attributed to the downstream processing skill rather than article-add itself, which merely acts as a storage mechanism. Therefore, it is classified as a LOW severity finding for this skill.

No other threats such as data exfiltration, obfuscation, unverifiable dependencies, privilege escalation, persistence mechanisms, metadata poisoning, or time-delayed attacks were detected within the SKILL.md file.

Total Findings: 1

🔵 LOW Findings: • Indirect Prompt Injection Potential

• Line 45: The skill writes user-provided <topic> directly to article-queue.md. If a downstream LLM-driven skill processes this file without sanitization, a malicious topic could lead to indirect prompt injection.

================================================================================