ml-workflow
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely instructional and descriptive, focusing on the architectural design of machine learning workflows. It does not perform any file system modifications, network requests, or command executions. Analysis of the content shows no evidence of prompt injection, obfuscation, or unauthorized data access.- [INDIRECT_PROMPT_INJECTION]: The skill's scope involves reading external files such as ML code, configuration files, and experiment logs. While this constitutes an attack surface for indirect prompt injection, the skill lacks the capabilities (network access, file writing, or command execution) necessary to act upon or exfiltrate data based on malicious instructions embedded in those files.
- Ingestion points: The skill reads ML code, configuration files, experiment logs, and infrastructure specifications for analysis.
- Boundary markers: No explicit delimiters or boundary markers are defined to separate ingested data from the agent's instructions.
- Capability inventory: None. The skill is designed to produce documentation and architecture diagrams without invoking shell commands or external tools.
- Sanitization: The procedure specifies that all inputs are treated as read-only analysis targets and no user-provided values are used in commands.
Audit Metadata