leaf-profile
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
leaf profilecommand to retrieve and merge configuration data from global and repository-specific profile files. - [PROMPT_INJECTION]: The skill manages a memory system that stores user instructions and requirements in profile files, which are later read and followed by the agent, creating a surface for indirect prompt injection.
- Ingestion points: Data is ingested from
~/.config/leaf/profile.mdand.leaf/PROFILE.mdvia theleaf profilecommand output. - Boundary markers: No explicit delimiters or instructions are used to distinguish stored profile content from current task instructions.
- Capability inventory: The skill has the capability to read and modify files in the user's configuration and repository directories via shell commands.
- Sanitization: There is no evidence of sanitization or validation performed on the content retrieved from or written to the profile files.
Audit Metadata