skills/jeremylongshore/tons-of-skills-marketplace/snowflake-data-quality-sentinel/Gen Agent Trust Hub
snowflake-data-quality-sentinel
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The script scripts/collect_snowflake_evidence.py uses subprocess.run to invoke the Snowflake CLI (snow). The command arguments, including the connection profile and SQL identifiers, are strictly validated against allowlist regular expressions to prevent command injection.
- [DATA_EXFILTRATION]: The skill includes comprehensive redaction and hashing logic. scripts/collect_snowflake_evidence.py redacts authorization headers, connection URLs, and sensitive keys from Snowflake CLI output. SQL results are hashed using SHA-256 before being processed to prevent the exposure of raw identifiers or values.
- [DYNAMIC_EXECUTION]: scripts/collect_snowflake_evidence.py performs string interpolation on SQL templates. To mitigate risks, the script implements validate_read_only_sql, which blocks a specific list of mutation and session-modifying SQL tokens (e.g., ALTER, DROP, GRANT, INSERT).
- [INDIRECT_PROMPT_INJECTION]: The skill ingests Snowflake metadata and metric history for analysis. 1. Ingestion points: scripts/analyze_data_quality.py reads JSON evidence from standard input or files. 2. Boundary markers: The analyzer enforces strict input contracts, including schema validation and independent SHA-256 digests for policy and evidence verification. 3. Capability inventory: scripts/collect_snowflake_evidence.py provides read-only Snowflake data collection via subprocess calls to the snow CLI. 4. Sanitization: The collector redacts sensitive authorization values and rejects credential-bearing fields before emitting results.
Audit Metadata