oracle
Warn
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
npx -y @steipete/oracle, which fetches a package from the npm registry. The-yflag ensures the package is installed and run without user confirmation. - [REMOTE_CODE_EXECUTION]: Executing a third-party package via
npxallows code from a remote repository to run on the local system. - [COMMAND_EXECUTION]: The skill requires the execution of shell commands, specifically
npxandhey re, to function and rebuild the agent's configuration. - [PROMPT_INJECTION]: The skill's 'Self-improvement' logic directs the agent to write new instructions to
~/.config/dotfiles/config/agents/rules/. These instructions are then automatically included in system prompts for future sessions, allowing the agent to persist and modify its own operating rules and constraints without explicit user review.
Audit Metadata