rpi-research
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFE
Full Analysis
- Structured Data Ingestion Guidelines: The skill contains explicit instructions to treat all fetched pages, repository files, and external content as inert data. It specifically directs the agent to ignore any instructions embedded within that data, such as prompt injection attempts or identity assertions, which is a key security measure for research-oriented agents.
- Read-Only Enforcement: The skill enforces a strict read-only boundary for the research phase. It restricts command execution to safe, data-gathering utilities like
git log,grep, andls, and prohibits the modification of source files, minimizing the risk of accidental or unauthorized alterations. - Path and Secret Security: There are clear constraints against recording credentials, tokens, or keys in artifacts. Additionally, the skill includes logic to reject relative path traversal (e.g.,
..) and absolute paths unless explicitly authorized, ensuring that research outputs are confined to designated tracking directories. - Instructional Context: The static analysis flags regarding 'ignoring instructions' refer to the skill's own internal security protocols which tell the agent to disregard instructions found inside the data it is researching (like a web page telling the AI what to do). This is a protective measure, not a malicious prompt injection.
Audit Metadata