who-are-you

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill gathers and displays system metadata that may be considered sensitive in some environments. Technical Metadata: Discloses the full physical path of the project (pwd -P) and the source IP address of the current SSH connection ($SSH_CONNECTION).
  • [PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by reading and reflecting content from local files and environment variables.
  • Ingestion points: The file CLAUDE.md is parsed for identity information, and environment variables are used to detect the terminal and shell.
  • Boundary markers: Content is reflected directly into the markdown output without specific delimiters or instructions to the model to ignore embedded commands.
  • Capability inventory: The skill uses bash to execute various system discovery commands.
  • Sanitization: Data from external files or the environment is not sanitized before being included in the output.
  • [COMMAND_EXECUTION]: The skill runs multiple shell commands to inspect the system state and configuration.
  • Command Details: Executes uname, pgrep, tmux, wezterm, and the project's default shell to retrieve version and session information.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 02:56 AM
Security Audit — agent-trust-hub — who-are-you