ljg-paper
Pass
Audited by Gen Agent Trust Hub on May 12, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE]: The skill is designed to manage academic notes within the user's
~/Documents/notes/directory. It reads from and writes to this specific path and accesses a local style guide at~/.claude/PAI/USER/AI_WRITING_PATTERNS.md. These operations are transparent and aligned with the skill's purpose as a personal note-taking assistant. - [EXTERNAL_DOWNLOADS]: The skill retrieves academic content from external URLs and Arxiv using the
WebFetchandWebSearchtools. This behavior is necessary for extracting data from research papers as requested by the user. - [PROMPT_INJECTION]: The skill processes external data (research papers) which creates an inherent surface for indirect prompt injection. However, the skill specifies a very rigid output structure (Org-mode template) and narrow extraction goals (Problem, Translation, Concepts, etc.), which naturally constrains the agent's behavior when processing untrusted input.
Audit Metadata