tidb-query-tuning
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill references a corpus of documentation derived from untrusted external content (public GitHub issues), which creates a surface for indirect instructions to bypass agent logic.
- Ingestion points: Multiple files within the
references/tidb-customer-planner-issues/directory contain content converted from raw user-submitted GitHub issues. - Boundary markers: Absent. There are no explicit markers or system-level instructions provided to the agent to treat this specific reference data as data only and ignore any embedded natural language instructions.
- Capability inventory: The skill has access to significant capabilities including SQL execution, subprocess execution for CLI tools like
tiupandgh, and local network access viacurlto TiDB debug APIs. - Sanitization: Absent. The data is mined from GitHub and converted to markdown without visible filtering for injection patterns.
- [DYNAMIC_EXECUTION]: The skill includes a Python script and instructs the agent on how to execute it to refresh its local knowledge base.
- Evidence:
scripts/generate_tidb_issue_experiences.pyis used to fetch data and generate markdown files at runtime.
Audit Metadata