ultrawork
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection where untrusted user input enters the agent context via the 'prompt' parameter in task delegation commands within SKILL.md. Boundary markers are absent, meaning malicious instructions in the task description could be obeyed by the subagent. Capability inventory includes full subprocess execution through the 'executor' subagent. No sanitization or filtering of external content is implemented.
- [COMMAND_EXECUTION]: The skill explicitly guides the agent to execute shell commands, including 'npm install' and 'npm run build', through delegated subagents. While this is a functional requirement for the task engine, it facilitates the execution of potentially arbitrary code derived from user-defined task prompts.
Audit Metadata