external-research
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting and acting upon external documentation.
- Ingestion points: Documentation is fetched from an external source referred to as 'Context7', cached locally in the
.tmp/external-context/directory, and then read by the agent. Additionally, user-supplied input is interpolated directly into the task via the$ARGUMENTSplaceholder at the end ofSKILL.md. - Boundary markers: There are no explicit instructions or delimiters provided to the agent to distinguish between the fetched documentation content and its core system instructions, nor are there warnings to ignore embedded commands within that data.
- Capability inventory: The agent is instructed to load these files and 'Apply to Implementation', meaning instructions or code patterns found in the external files will directly influence the agent's code generation and file-writing capabilities.
- Sanitization: The skill lacks any description of sanitization, filtering, or validation for the content fetched from the external documentation source before the agent processes it.
Audit Metadata