invention-structuring
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-provided arguments and local patent-related files, which are then interpolated into its internal logic and reviewer prompts without sufficient isolation or sanitization.
- Ingestion points: Data is ingested from the
$ARGUMENTSvariable and from files:patent/INVENTION_BRIEF.md,patent/PRIOR_ART_REPORT.md, andpatent/NOVELTY_ASSESSMENT.md. - Boundary markers: The instructions do not use delimiters (e.g., XML tags or block markers) or provide explicit warnings to the model to ignore instructions embedded within the ingested data.
- Capability inventory: The skill is configured with powerful tools including
Bash(*),Write, andEdit, creating a potential impact if malicious instructions in the data are executed. - Sanitization: No input validation or filtering is performed on the data before it is used to drive the invention decomposition and validation workflow.
- [COMMAND_EXECUTION]: The workflow involves the execution of a local script named
reviewer_route.pyto resolve reviewer identities and route requests. This instruction directs the agent to execute a file-based dependency to manage its operational logic.
Audit Metadata