yadm
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
yadmCLI commands to manage system configuration files. This includes the ability to run theyadm bootstrapscript, which is a standard procedure for environment setup but involves executing shell scripts located in the user's dotfiles repository. - [DATA_EXFILTRATION]: The skill is designed to sync local configuration files with a remote repository using
yadm push origin main. While this is the intended functionality of a dotfile manager, it inherently involves transferring local data to an external Git server. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from the local file system (dotfiles) and repository metadata.
- Ingestion points: Commands like
yadm status,yadm diff, andyadm listread content from the local workspace. - Boundary markers: The instructions state the agent should remain read-only unless a modification is explicitly requested.
- Capability inventory: The agent can perform file writes, commits, and pushes to remote repositories, and execute the bootstrap shell script.
- Sanitization: The skill instructs the agent to use
--stator--name-statusfor sensitive files to avoid echoing secret contents into the session logs. - [EXTERNAL_DOWNLOADS]: The skill references the official source code for the
yadmtool hosted on GitHub for documentation purposes.
Audit Metadata