agent-architecture

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses lifecycle hooks to execute shell commands for session logging and state management. Evidence: 'pre' and 'post' hooks in SKILL.md executing echo, memory_store, and memory_search commands.
  • [PROMPT_INJECTION]: The skill ingests content from previous phases via memory retrieval, representing an indirect injection surface. 1. Ingestion points: 'memory_search' in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Platform-specific shell hooks. 4. Sanitization: Absent; content is sourced from internal memory stores.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 11:06 AM
Security Audit — agent-trust-hub — agent-architecture