reviewing-ml-prior-art-survey
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a structured framework for an AI agent to act as a quality gate for ML prior-art surveys. It defines clear boundaries for evidence and requires findings to be grounded in specific project-local files.
- [INDIRECT_PROMPT_INJECTION]: The skill correctly identifies that the data it reviews (model cards, third-party documentation) may contain untrusted content. It explicitly includes Condition C21, which instructs the reviewer to ensure no instructions found in fetched content were followed, effectively treating third-party content as untrusted data.
- [DATA_EXPOSURE_AND_EXFILTRATION]: While the skill references configuration files such as
source-registry.yamland various schemas, these are project-specific artifacts required for the review process. There is no evidence of attempts to access sensitive system files, environment variables, or exfiltrate data to external domains. - [COMMAND_EXECUTION]: The skill does not contain any shell commands, subprocess invocations, or scripts. It operates entirely as a set of logical instructions for the AI agent.
Audit Metadata