ai-memory-routing-install

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to perform file system operations, including writing, refreshing, and removing files in sensitive locations such as project-local directories (.claude/skills/, .agents/skills/) and global user directories (~/.claude/skills/, ~/.agents/skills/, ~/.grok/skills/). While these are intended targets for agent configuration, unauthorized or accidental modification of these paths could impact agent behavior.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external data to update agent instructions.
  • Ingestion points: The skill ingests instruction snippets and skill payloads returned by the memory_install_self_routing tool.
  • Boundary markers: It utilizes specific markers (<!-- ai-memory:start -->, <!-- ai-memory-managed: routing-skill -->) to delimit managed content.
  • Capability inventory: The skill possesses capabilities to write and edit files in local and global configuration roots across multiple agent platforms.
  • Sanitization: The instructions rely on marker-based replacement and platform-aware path joining but do not explicitly detail sanitization or validation of the content payloads provided by the tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 01:22 PM
Security Audit — agent-trust-hub — ai-memory-routing-install