deeplearning-ai-agentic-research-agent
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches code from an external GitHub repository (
github.com/https-deeplearning-ai/agentic-ai-public.git). This repository and organization are not established trusted sources, posing a supply chain risk. - [INDIRECT_PROMPT_INJECTION]: The skill implements a multi-agent workflow that ingests content from untrusted external sources and interpolates it directly into LLM prompts.
- Ingestion points: User input (
prompt), results fromtavily_search_tool(web results),arxiv_search_tool(academic papers), andwikipedia_search_tool(Wikipedia summaries). - Boundary markers: Absent. The skill uses simple string interpolation (f-strings) to combine untrusted data into system and user messages for the agents.
- Capability inventory: The agents can perform network requests to various APIs and write task state/reports to a Postgres database.
- Sanitization: None. The content is passed raw to the LLM for synthesis and drafting.
- [COMMAND_EXECUTION]: The installation and execution instructions require the user to run shell commands (
git clone,docker build,docker run) and execute a containerized service that performs automated background tasks.
Audit Metadata