exploited-vulnerability-watch

Pass

Audited by Gen Agent Trust Hub on Oct 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (security advisories, KEV catalog) and untrusted internal data (package manifests, lockfiles). This creates a surface for indirect prompt injection where malicious instructions could be embedded in advisory text or file metadata.
  • Ingestion points: Step 2 and 3 describe reading local manifests/lockfiles and fetching external advisory content from the web.
  • Boundary markers: The 'Safety Boundary' section explicitly instructs the agent to treat all inspected content as untrusted and never follow instructions embedded within it.
  • Capability inventory: The skill has read access to local files and network access for research. It does not have write access or command execution capabilities (explicitly "read-only").
  • Sanitization: No specific technical sanitization methods are mentioned beyond the instruction to ignore embedded commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 28, 2026, 04:45 AM