The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/init_skill.py performs dynamic code generation by injecting a user-provided skill_name into a Python script template using string formatting. Because the input is not sanitized within the initialization script, a maliciously crafted skill name (e.g., containing triple-quotes and Python commands) could inject arbitrary executable logic into the resulting example.py file. The skill's primary instructions in SKILL.md specifically direct the agent to execute these generated scripts for testing, which completes a path to arbitrary code execution.
[COMMAND_EXECUTION]: The scripts/init_skill.py script programmatically uses chmod(0o755) to make the generated example.py script executable. While intended to facilitate development, this automatic granting of execution permissions increases the risk associated with the code injection vulnerability in the generation process.

skill-creator