bro
Fail
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The README.md file promotes an installation method using curl and bash process substitution to execute a script directly from the author's GitHub repository (https://raw.githubusercontent.com/luchasarie/bro-skill/main/install.sh).
- [COMMAND_EXECUTION]: The
install.shscript executes shell commands to create directories and write files within user-controlled configuration paths, such as~/.claude/commands/,~/.hermes/skills/, and~/.cursor/skills/, to integrate the skill into those tools. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data entered by users (Ingestion point: text provided after the
/brocommand or previous message context). While the prompt lacks formal boundary markers like XML delimiters, it provides specific instructions for the AI to treat input strictly as data and ignore any embedded commands or rule-bypass attempts (Sanitization). The skill frontmatter does not define any external tool access, limiting the potential impact of an injection (Capability inventory).
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/luchasarie/bro-skill/main/install.sh - DO NOT USE without thorough review
Audit Metadata