The Agent Skills Directory

[PROMPT_INJECTION]: The scripts/generate_schema.py script interpolates user-supplied database descriptions directly into an LLM prompt (user_prompt) without using boundary markers (such as XML tags or triple quotes) or input sanitization. This creates an indirect prompt injection surface where a malicious description could influence the generated SQL or code.
Ingestion points: The description argument in scripts/generate_schema.py receives untrusted user input.
Boundary markers: Absent. The input is concatenated directly into the prompt string.
Capability inventory: The skill possesses the ability to execute arbitrary SQL on a target database using the execute_sql tool in the Supabase MCP server (invoked via scripts/apply_schema.py).
Sanitization: Although scripts/validate_schema.py checks for basic SQL syntax and best practices, it does not specifically filter for malicious SQL patterns or prompt injection artifacts.
[COMMAND_EXECUTION]: The scripts/apply_schema.py script utilizes subprocess.run() to execute local Python scripts and the manus-mcp-cli platform tool. While these calls use argument lists which mitigate shell injection risks, they enable the skill to interact with the system's command line and apply generated SQL content to live database environments.

database-schema-generator