The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to configure Model Context Protocol (MCP) servers using npx and docker. These commands target official vendor resources, specifically @elastic/mcp-server-elasticsearch on the NPM registry and docker.elastic.co/mcp/elasticsearch on the Elastic container registry. These are recognized as trusted vendor resources.
[REMOTE_CODE_EXECUTION]: Setup instructions in references/mcp-setup/mcp-setup.md involve running remote code via npx and docker run. As these target official Elastic-maintained packages, they are considered safe under the trusted source guidelines but are noted as execution patterns.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret user-provided data (samples, schemas, and descriptions) to assist in creating Elasticsearch mappings, ingestion pipelines, and search queries. This creates a surface for indirect prompt injection where malicious data could attempt to influence the agent's logic.
Ingestion points: Data samples (JSON/CSV) and schema descriptions provided by the user during the onboarding flow in references/elasticsearch-onboarding-playbook.md.
Boundary markers: The skill employs a strict 'Write Confirmation Protocol' (detailed in references/code-generation/code-generation.md and references/mcp-setup/mcp-setup.md), which requires the agent to show the exact API call and obtain explicit user approval before execution.
Capability inventory: The skill can generate and suggest code in multiple languages (Python, JS, Go, etc.) and perform cluster management tasks via the Elasticsearch MCP tools.
Sanitization: The agent is instructed to explain API patterns before implementation and wait for user confirmation, serving as a manual validation step for any processed data.

elasticsearch-onboarding