skills/edwinhu/workflows/paperpile/Gen Agent Trust Hub

paperpile

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of local CLI tools (paperpile, scholar, bun) and shell scripts to perform library management and PDF resolution. These are standard for the tool's documented purpose.
  • [EXTERNAL_DOWNLOADS]: The skill downloads academic papers and metadata from well-known scholarly sources including CrossRef, SSRN, and various university library proxies (UVA, NYU). These operations are transparently documented as part of the core PDF resolution functionality.
  • [COMMAND_EXECUTION]: The script scripts/resolve_pdf.py uses subprocess.run to call curl and other local utilities for network requests and metadata lookups. Static analysis correctly identified these calls; however, they are used with hardcoded or sanitized parameters and do not pose an injection risk in this context.
  • [REMOTE_CODE_EXECUTION]: The skill interacts with Chrome via the Chrome DevTools Protocol (CDP) on port 9250 to automate browser-based PDF downloads and session warmup. This is a functional requirement for bypassing institutional SSO challenges and is not used for arbitrary code execution.
  • [CREDENTIALS_UNSAFE]: The skill manages authentication by snapshotting and hydrating browser cookies into local JSON files at ~/.claude-work/skills/paperpile/cookies/. While these files contain session data, they are stored within the user's local config directory with limited file permissions (chmod 600), following standard practice for CLI authentication persistence.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 06:29 AM
Security Audit — agent-trust-hub — paperpile