command-creator
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEPERSISTENCEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PERSISTENCE]: The skill automates the creation of markdown files in the global
~/.claude/commands/directory. These files are persistent across Claude Code sessions and are automatically loaded as executable slash commands, allowing for the creation of persistent automated workflows.\n- [DYNAMIC_EXECUTION]: The skill is designed to generate executable agent instruction sets (slash commands) based on user input and pre-defined patterns. This capability allows for the generation of complex, autonomous automation scripts that increase the potential impact of malicious or poorly crafted instructions.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input for command names, descriptions, and workflow steps, which are then interpolated into new command files without explicit sanitization or validation. This represents an injection surface where user-supplied content could subvert the intended behavior of the generated command.\n - Ingestion points: Step 3 of SKILL.md gathers user input for the command name, description, and execution steps.\n
- Boundary markers: The generated output follows a markdown template but lacks strict structural enforcement to prevent content escaping from intended fields.\n
- Capability inventory: The skill uses the Write tool to create files and the Bash tool to manage directories.\n
- Sanitization: No explicit sanitization or filtering of user input is performed before file creation.\n- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute directory creation commands (
mkdir -p) when setting up the command storage environment.
Audit Metadata