prior-art-search
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from various sources, which could contain malicious instructions intended to influence the agent's behavior.
- Ingestion points: The skill reads input from
$ARGUMENTS, local files (patent/INVENTION_BRIEF.md,INVENTION_BRIEF.md), shared reference files, and results fromWebSearchandWebFetch(Google Patents, Espacenet, Google Scholar, arXiv). - Boundary markers: There are no explicit instructions or delimiters defined to isolate user-provided text or web content from the agent's core instructions.
- Capability inventory: The agent has access to
Bash(*),Write,WebSearch, andWebFetchtools. - Sanitization: There is no evidence of sanitization, filtering, or validation of the technical descriptions or web content before processing.
- [COMMAND_EXECUTION]: The skill configuration enables full access to the Bash tool via
allowed-tools: Bash(*). - Evidence: The workflow instructs the agent to perform analysis and generate reports. While the instructions do not explicitly call for dangerous commands, the combination of processing untrusted input data with a general-purpose shell tool creates a surface for potential command injection if the input content successfully manipulates the agent's output into shell commands.
Audit Metadata