research-knowledge-graph
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill performs network requests to
api.openalex.orgto enrich bibliographic data. These requests are standard for the tool's primary purpose and target a well-known scientific metadata provider. It correctly implements a rate-limiting sleep of 0.1 seconds and provides a--mailtoparameter for the API's polite pool, which is considered a best practice for this service. - [COMMAND_EXECUTION]: The
SKILL.mdfile provides examples of command-line execution for the provided Python script. These commands are standard for local tool operation and do not involve suspicious behavior like piping remote scripts to a shell. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted bibliographic data (
.bibor.jsonfiles) which could contain embedded instructions. However, thebuild_graph.pyscript uses a custom parser for BibTeX and standard JSON parsing to extract specific fields into structured graph data. The risk is minimized as the data is primarily used to build graph structures and Markdown files rather than being directly interpolated into an LLM prompt without sanitization. The skill instructions also include explicit integrity rules for the agent, such as 'Never invent edges or papers'.
Audit Metadata