wave-execute

Pass

Audited by Gen Agent Trust Hub on Aug 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted data from story files (e.g., s1.md, path1.md) which are used to drive sub-agent tasks and CLI parameters.
  • Ingestion points: Story paths resolved in Stage 1 and read during Stage 3 dispatch.
  • Boundary markers: The skill does not define clear boundary markers or delimiters for story file content processed by sub-agents.
  • Capability inventory: The skill executes 'aiox wave' and 'aiox sdc' commands via shell and spawns sub-agents ('aiox-master') with the authority to plan and verify work.
  • Sanitization: Stage 1 involves 'validating budget' and 'scanning child intents,' which provides a check before model invocation, but does not fully sanitize the content of the story files being read.
  • [COMMAND_EXECUTION]: The skill relies on the 'aiox' CLI utility suite to manage state and execute the wave lifecycle.
  • Commands include 'aiox wave plan', 'aiox wave advance', 'aiox wave next', 'aiox wave mark', and 'aiox wave status'.
  • The skill also spawns sub-agents with instructions to run 'aiox sdc plan/next/verify' within the target story directory.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 9, 2026, 03:07 AM
Security Audit — agent-trust-hub — wave-execute