commercialize-academic-research
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute local helper scripts (scripts/init_case_pack.py, scripts/add_source_log.py, scripts/validate_case_pack.py) to create and verify directory structures and markdown templates for commercialization analysis. These scripts perform standard local filesystem operations (directory creation and file writing) using standard Python libraries without any network activity or privilege escalation.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted research assets such as papers or abstracts. This creates a surface for indirect prompt injection if the asset contains adversarial instructions.
- Ingestion points: The 'research_asset' input field in SKILL.md serves as the ingestion point for untrusted academic text.
- Boundary markers: The skill provides structural guidance (Analysis chain, Commercialization Evidence Table) to ground the agent's output, although it does not utilize explicit delimiters for the asset content.
- Capability inventory: The skill allows the agent to write files to the local filesystem via the provided Python scripts.
- Sanitization: The 'init_case_pack.py' script normalizes input names for directory creation, but the LLM-processed content depends on the agent's internal safety filters.
Audit Metadata