keboola-context
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and process data from the Keboola telemetry database (Snowflake) to generate summary statistics and PR comments. This creates a vulnerability surface where untrusted data within the telemetry records could influence the agent's behavior or output.
- Ingestion points: Data is retrieved from Snowflake tables (e.g.,
kbc_component_configuration,kbc_job) via thequery_datatool in thekeboola-mcp-us-east4gcpMCP server. - Boundary markers: No explicit delimiters (e.g., XML tags or backticks) are specified for the ingested database content in the prompt instructions.
- Capability inventory: The agent is instructed to use the MCP server to perform SQL queries and subsequently write comments to pull requests.
- Sanitization: The skill provides a comprehensive "Anonymization Rules" table in
references/telemetry.mdwhich explicitly forbids including identifying information like Project IDs, client names, or organization names in output, acting as a manual sanitization filter.
Audit Metadata