The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from multiple external sources including Notion, dbt, Looker, Metabase, and local markdown wikis to generate context for AI agents.
Ingestion points: Data is imported from dbt manifest files, Looker projects, Notion pages, and local markdown files in the wiki/ directory as described in SKILL.md.
Boundary markers: The documentation does not specify the use of delimiters or specific instructions to the agent to ignore embedded commands within the ingested context data.
Capability inventory: The skill provides MCP tools such as ktx_search_semantic_layer and ktx_search_wiki which allow agents to interact with and query information from data warehouses.
Sanitization: There is no mention of sanitization, filtering, or validation processes for the data ingested from external sources.
[COMMAND_EXECUTION]: The skill documentation includes code examples for interacting with the tool via TypeScript that use potentially unsafe command execution patterns.
Evidence: The "Querying from TypeScript" section provides an example using child_process.exec to run ktx sl "${query}", which creates a command injection surface if the query input is not properly sanitized.
[EXTERNAL_DOWNLOADS]: The skill instructions require downloading and installing the @kaelio/ktx package and potentially adding skills from external repositories.
Evidence: The skill requires npm install -g @kaelio/ktx for global installation and suggests using npx skills add Kaelio/ktx --skill ktx to integrate with Claude Code.

ktx-ai-data-agents-context-layer