create-integration

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill automates the installation of the ceplx tool from github.com/efd6/ceplx using go install. This repository is a third-party resource rather than an official organization account, which introduces a potential supply chain risk if the external repository were compromised.
  • [EXTERNAL_DOWNLOADS]: The workflow involves fetching and installing multiple Go-based utilities. It neutrally downloads the mito, celfmt, and stream tools from Elastic's official GitHub organization to provide necessary functionality for integration building. It also downloads the ceplx tool from a third-party repository.
  • [COMMAND_EXECUTION]: The instructions direct the agent to execute various shell commands to manage the environment and build integration packages, including go install, elastic-package check, elastic-package stack up, and celfmt.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, including API documentation URLs, research briefs, and sample event logs. While the skill includes instructions for data anonymization, the lack of explicit boundary markers for this ingested content creates a surface where embedded instructions could influence the agent's behavior during the build process.
  • Ingestion points: Files provided via @-mentions (e.g., @samples/acme_event.json) and external documentation fetched from URLs.
  • Boundary markers: None explicitly defined to separate untrusted data from instructions.
  • Capability inventory: The skill has file-write permissions for scaffolding, shell command execution via elastic-package, and the ability to delegate tasks to subagents.
  • Sanitization: Mandates that all data committed to the repository must be fully anonymized using synthetic values.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 10:59 AM