configuring-ai-agents
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists of instructional content and OpenAPI-compliant schema definitions for the Celigo platform. All identified commands and URLs are standard for integration and AI agent management tasks.
- [INDIRECT_PROMPT_INJECTION]: As a configuration guide for AI agents, the skill describes a system designed to process potentially untrusted input through LLMs. While this represents the primary purpose of the skill, it is noted as an inherent attack surface.
- Ingestion points: Data mapped from integration records to agent input fields (text, record, files) as defined in SKILL.md.
- Boundary markers: The skill emphasizes 'Prompt design' for behavior control but does not mandate specific injection delimiters in its examples.
- Capability inventory: Supports high-capability tools including web search, MCP server connections, and provider-native code execution as documented in SKILL.md.
- Sanitization: Leverages structured output formats like json_schema to improve reliability and downstream parsing.
Audit Metadata