skills/quantipixels/skills/pese/Gen Agent Trust Hub

pese

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes and serves local file content, which constitutes an attack surface for indirect prompt injection if the files contain malicious instructions.\n
  • Ingestion points: Local files or directories provided by the user as targets for sharing (SKILL.md).\n
  • Boundary markers: The skill emphasizes pinning specific resources but does not define delimiters for the content of the files themselves.\n
  • Capability inventory: The skill utilizes network tunneling tools (Tailscale, Tailcat) and may perform software installation if tools are missing.\n
  • Sanitization: No sanitization is performed on the contents of the files being served, as the file content is served as-is.\n- [EXTERNAL_DOWNLOADS]: The skill directs users to official documentation and repositories for Tailscale and Tailcat for installation and usage guidance.\n
  • Evidence: References official domains (tailscale.com) and repositories (github.com/tailscale/tailcat). These are well-known technology services.\n- [DATA_EXFILTRATION]: The skill's primary function is to make local data accessible over a network. While scoped to private transport (tailnets), this is a data exposure mechanism.\n
  • Mitigation: The instructions explicitly mandate ephemeral server identities, rejection of sensitive directories (like .ssh or credentials), and immediate revocation after use to minimize risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 05:39 PM
Security Audit — agent-trust-hub — pese