The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_migration_script.py contains a generate_script method that constructs Bash scripts (.sh) by interpolating input content into a shell template. This runtime generation of executable scripts from string data is a high-risk pattern as it provides a mechanism for arbitrary command construction.
[COMMAND_EXECUTION]: The generate_migration_script.py utility explicitly executes file_path.chmod(0o755) on newly created scripts. Making dynamically generated files executable is a sensitive operation that can be leveraged to bypass security controls and facilitate the execution of potentially malicious code.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted schema data from external databases and using it to influence the logic of generated migration scripts.
Ingestion points: External database objects and metadata retrieved via pg_dump, mysqldump, or direct SQL queries against information_schema.
Boundary markers: None identified; the skill does not use delimiters or instructions to prevent the agent from obeying commands embedded within database comments or object names.
Capability inventory: The skill has the ability to write to the local file system, modify file permissions, and execute shell commands through the Bash tool.
Sanitization: There is no implementation of escaping or validation logic to sanitize schema definitions before they are used to build SQL or shell scripts.

comparing-database-schemas