ktx-ai-data-agents-context-layer
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill facilitates the installation of the @kaelio/ktx package from the NPM registry. This package is the core component of the ktx tool provided by the skill author.
- [COMMAND_EXECUTION]: The skill executes various ktx CLI commands, such as ktx setup, ktx ingest, and ktx mcp start, to manage the semantic layer and interface with AI agents. These commands are part of the intended local workflow.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill documentation guides the user in setting up environment variables for API keys and database credentials. It correctly recommends using .env files for local secret management and highlights that the tool uses read-only database operations.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests data from external sources like data warehouses, Notion, and wikis to build a context layer for AI agents. Ingestion points: Data is ingested via ktx ingest from databases, dbt manifests, and markdown wiki files. Boundary markers: No explicit boundary markers or instruction-ignoring delimiters are specified for the ingested content. Capability inventory: The skill provides an MCP server that exposes semantic search and metric retrieval tools to AI agents. Sanitization: No sanitization or validation of the external metadata or wiki content is mentioned in the instructions.
Audit Metadata