customizing-commands

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides guidance for writing and saving TypeScript files to the ~/.letta/mods/ directory. These scripts are intended to be registered and executed by the application to provide custom slash command functionality.
  • [PROMPT_INJECTION]: The skill includes code templates that interpolate raw user input from command arguments into agent prompts (e.g., Review current git changes. Focus on ${focus}.). This creates a vulnerability surface for indirect prompt injection as it lacks boundary markers or input sanitization.
  • Ingestion points: User-supplied arguments in ctx.args (SKILL.md).
  • Boundary markers: Absent from prompt templates.
  • Capability inventory: Execution of arbitrary logic within registered mod scripts.
  • Sanitization: None provided in the implementation examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 10:42 PM
Security Audit — agent-trust-hub — customizing-commands