paseo-orchestrator

Warn

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on the paseo CLI to perform sensitive operations, including paseo run with elevated privileges (--mode bypassPermissions) and paseo schedule create for establishing persistent heartbeat tasks.
  • [PROMPT_INJECTION]: The orchestrator is vulnerable to indirect prompt injection because agents process untrusted data from chat rooms and the codebase. (1) Ingestion points: Agents read full histories from chat rooms via paseo chat read. (2) Boundary markers: No delimiters or instructions to ignore embedded commands are present in the logic. (3) Capability inventory: Agents can execute sub-agents with full access and bypassed permissions. (4) Sanitization: No validation or escaping of external chat content is implemented before it influences agent behavior.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 4, 2026, 02:51 PM