edge-computing

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute directory listing commands (ls) on specific hidden paths (e.g., ~/.claude/skills/) to check for the presence of recommended companion skills in the environment.
  • [COMMAND_EXECUTION]: The instructions suggest using npx skills add to install additional components from the vendor's repository, which involves executing code from an external registry.
  • [PROMPT_INJECTION]: The skill contains logic for fetching and processing external content from origin servers (e.g., modifying HTML responses and streaming data). This creates a surface for indirect prompt injection where malicious instructions embedded in the fetched origin content could influence the agent's behavior.
  • Ingestion points: Origin fetches in SKILL.md (e.g., await fetch(request)) and references/latency-optimization.md.
  • Boundary markers: Absent in instructions and code examples.
  • Capability inventory: Includes network access (fetch), HTML string manipulation, and response streaming.
  • Sanitization: No sanitization or validation of the fetched content is provided in the examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 01:04 PM
Security Audit — agent-trust-hub — edge-computing