The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and adapt user-provided SQL code for TiDB compatibility. This represents a potential surface for indirect prompt injection.
Ingestion points: User-supplied SQL queries provided to the agent for review, migration, or debugging (SKILL.md).
Boundary markers: The instructions lack explicit directives for using boundary markers or delimiters when handling untrusted SQL input.
Capability inventory: The skill utilizes agent tools to execute SQL queries and includes a local shell script (render_dot_png.sh) for DOT file rendering.
Sanitization: No sanitization or validation protocols are established for the input SQL data.
[COMMAND_EXECUTION]: The skill includes a shell script, scripts/render_dot_png.sh, which facilitates the visualization of TiDB query plans using the dot utility. The script employs standard shell scripting best practices, such as variable quoting, to minimize the risk of command injection.
[SAFE]: The documentation in references/tidb-cloud-ssl.md provides essential security instructions for connecting to TiDB Cloud, emphasizing the use of SSL with certificate and identity verification (--ssl-mode=VERIFY_IDENTITY). This guidance effectively addresses risks associated with data intercept in transit.

tidb-sql