citation-graph-ingest
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process third-party document corpora, such as law, academic papers, and patents, which creates an attack surface where malicious instructions could be embedded in the processed data.
- Ingestion points: External document content is retrieved using
gbrain searchandgbrain getinSKILL.md(Pipeline steps 1 and 2). - Boundary markers: The skill explicitly references untrusted-content conventions and instructs the agent to treat citation text as DATA, specifically warning that imperatives embedded in a document must not decide the edge type.
- Capability inventory: The skill uses
gbrain linkto write typed edges to the native links table andgbrain graph-queryto traverse the graph. - Sanitization: The skill relies on model judgment and explicit instructions to disregard embedded imperatives within the corpus, focusing only on the actual citation context.
Audit Metadata