claude-code-hook-development

Pass

Audited by Gen Agent Trust Hub on Apr 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the creation of shell scripts in the .claude/hooks/ directory and configuring the agent to execute them via .claude/settings.json. This is the documented mechanism for Claude Code hooks and is used here for tasks like blocking dangerous commands, enforcing linting, and running tests.
  • [PROMPT_INJECTION]: The 'Prompt-Based Hooks' section in references/hook-events.md describes a potential surface for indirect prompt injection.
  • Ingestion points: Untrusted data enters the agent context via the $ARGUMENTS variable in prompt templates.
  • Boundary markers: The provided example (Evaluate if this is safe: $ARGUMENTS) lacks explicit delimiters or instructions to ignore embedded commands.
  • Capability inventory: Hooks have the capability to execute bash scripts and modify tool inputs across the system.
  • Sanitization: The examples do not demonstrate sanitization or validation of the interpolated arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 28, 2026, 08:49 AM
Security Audit — agent-trust-hub — claude-code-hook-development