ulw-loop
Warn
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The bootstrap logic in
references/full-workflow.mdexecutes a shell script that searches for and aliases a JavaScript executable from local paths such as plugin caches. This dynamic execution of scripts from variable local directories is a security risk. - [PROMPT_INJECTION]: The skill is susceptible to command injection through the user-provided 'brief' parameter. Because this input is interpolated directly into a shell command line (
omo ulw-loop create-goals --brief "<brief>"), a malicious user could execute arbitrary code by including shell metacharacters in their request. - [EXTERNAL_DOWNLOADS]: The instructions recommend installing the 'lazycodex-ai' package via
npx, which executes code from an unverified third-party registry. - [PROMPT_INJECTION]: The skill lacks sanitization for user input used in sub-agent orchestration. The 'brief' data is passed into the messages of spawned agents, creating a risk for indirect prompt injection where a user can manipulate the behavior of child agents.
- [SAFE]: The skill references 'agent-browser' from Vercel Labs' GitHub repository, which is considered a trusted source.
Audit Metadata