mem-sync
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a Bash script (mem-sync-git.sh) that performs several Git-related operations, including git worktree, git push, git fetch, and git rebase. These are core to the skill's synchronization functionality.
- [EXTERNAL_DOWNLOADS]: The script uses git fetch to retrieve daily log updates from remote Git repositories to synchronize state between different machines.
- [DATA_EXFILTRATION]: The skill is designed to push local data from the .memories/ directory to a remote Git repository. While this is intentional behavior for cross-device syncing, users should verify the configured sync remote is a trusted destination.
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it processes and outputs the contents of synchronized logs. Ingestion points: The status and diff commands in mem-sync-git.sh read and display the content of files from the .memories/ folder. Boundary markers: There are no explicit markers or warnings used to separate the log content from the agent's instructions. Capability inventory: The skill possesses significant command execution capabilities via shell and Git, which could be exploited if an agent follows instructions hidden within log files. Sanitization: No sanitization or filtering of the log file content is performed before it is presented to the agent.
Audit Metadata