anfra-publish
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use several shell commands including
anfra publish,anfra pull,anfra validate,git status, andgit cloneto manage repository state and synchronization. - [INDIRECT_PROMPT_INJECTION]: The skill processes repository content (models, datasets) and configuration files such as
.anfra/config.ymland.anfra/data_sources.ymlwhich could contain untrusted data. - Ingestion points: Reads repository files for conflict resolution and validation as described in
SKILL.md. - Boundary markers: The skill relies on standard conflict markers (
<<<<<<<,=======,>>>>>>>) to identify conflicting regions in file content. - Capability inventory: The skill utilizes
anfra publish,anfra pull,anfra validate, andgit clonefor operations across all scenarios inSKILL.md. - Sanitization: No explicit sanitization or filtering of file content is mentioned, though the instructions mandate manual user review and approval for all conflict resolutions.
Audit Metadata