volume-bootstrap

Pass

Audited by Gen Agent Trust Hub on Jun 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists of instructional and reference documentation regarding sandbox initialization and persistence contracts. It does not contain executable code or instructions that bypass safety guidelines.
  • [PROMPT_INJECTION]: The skill documents the ingestion of external data via tools like load_document and search_knowledge. This is a standard platform capability. The documentation notes that security is maintained through path traversal rejection and canonical root validation.
  • Ingestion points: External documents stored in knowledge/ingested/ (processed in SKILL.md).
  • Boundary markers: Documentation explicitly states that path traversal sequences like .. or URL-encoded variants are rejected.
  • Capability inventory: Capabilities include persistent memory writes (remember), file system writes to artifacts/, and session buffer management.
  • Sanitization: Input is validated against VFS_CANONICAL_ROOTS to prevent unauthorized file access.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 20, 2026, 02:55 PM
Security Audit — agent-trust-hub — volume-bootstrap