sqlmesh
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill facilitates the generation and execution of SQL and Python code (models and macros). This architecture is susceptible to indirect prompt injection if the agent processes untrusted inputs or model files containing malicious instructions.\n
- Ingestion points: Data transformation models in
models/, Python logic inmacros/, and external data retrieved via database adapters.\n - Boundary markers: Absent. There are no instructions to delimit or treat external code/data as untrusted.\n
- Capability inventory: SQL execution, Python macro execution, and environment state management via the
sqlmeshCLI.\n - Sanitization: Not implemented within the skill's instructions.\n- [COMMAND_EXECUTION]: The skill utilizes the
sqlmeshCLI to perform actions like initializing projects, planning changes, and applying backfills. These commands are necessary for the skill's intended purpose.\n- [EXTERNAL_DOWNLOADS]: The skill recommends installing thesqlmeshpackage and its various adapter extras (e.g., Spark, BigQuery) from the standard Python package registry.
Audit Metadata