The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the cockroach CLI tool, specifically cockroach sql --url "<url>" -e "SQL". This allows for direct system interaction and database execution, which could be exploited if the agent constructs commands using unsanitized user inputs.\n- [INDIRECT_PROMPT_INJECTION]: The skill converts natural language from users into executable SQL queries, presenting a clear injection surface. \n
Ingestion points: Natural language questions and intent provided in user prompts.\n
Boundary markers: None defined in the instructions to separate user input from the generated SQL logic.\n
Capability inventory: Execution of database writes (INSERT, UPDATE, DELETE) and schema modifications via shell execution.\n
Sanitization: The skill mandates syntax validation using EXPLAIN, which does not prevent malicious query logic or data modification.\n- [CREDENTIALS_UNSAFE]: The skill instructs the agent to access and echo the COCKROACH_URL environment variable to identify connection strings. These strings often contain sensitive plaintext credentials (username, password, and host information) that are exposed during this process.

cockroachdb-sql