claude-hooks
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent on how to configure and execute arbitrary shell commands using the
commandhook type in~/.claude/settings.json. - [INDIRECT_PROMPT_INJECTION]: The provided hook templates interpolate environment variables such as
$CLAUDE_MESSAGE,$CLAUDE_TOOL_INPUT, and$CLAUDE_TOOL_OUTPUTdirectly into shell commands. This creates a vulnerability where malicious data from external sources (e.g., a web page read by the agent or a notification message) could lead to command injection on the user's system. - Ingestion points: Environment variables populated from tool interactions and notifications, as documented in
references/claude-hook.template.md. - Boundary markers: None present; the templates do not use delimiters or instructions to ignore embedded commands within the variables.
- Capability inventory: The skill facilitates the execution of shell commands through the hook configuration system.
- Sanitization: No evidence of escaping, validation, or filtering of the environment variables before they are used in shell commands.
Audit Metadata