iblai-api-agent-dataset
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides an interface to ingest untrusted external data (URLs, website crawls, GitHub repositories, and files) into an agent's Retrieval-Augmented Generation (RAG) system. While this is the intended functionality, it creates a surface for indirect prompt injection if the ingested content contains malicious instructions.
- Ingestion points: Resource addition endpoints in
SKILL.md(URL, YouTube, Blackboard, Website crawl, GitHub, and File). - Boundary markers: Not specified within the skill instructions; handled by the platform's training pipeline.
- Capability inventory: Network operations restricted to vendor API and GitHub.
- Sanitization: The skill enforces strict validation for
custom_metadata(alphanumeric keys, scalar values), providing some structural integrity, though the content of the training resources is ingested as-is. - [SAFE]: The skill interacts exclusively with the vendor's official domain (
api.iblai.app) and the well-known GitHub API (api.github.com). It correctly utilizes environment variables for secret management and includes a mandatory user confirmation step for data deletion.
Audit Metadata