AI agents use git_restore to create or update resources in Git — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Git environment.
git_restore modifies files and the index by restoring them to a previous state. While this is reversible (you can restore again to a different state), it does overwrite local changes. It is not destructive in the sense of permanent deletion (that would be git_clean with -f). It's more severe than a Read operation because it alters the local state, but less severe than Destructive operations like deletion.
From the tool's definition Tool description states it can 'Restore paths in worktree and/or index from current state or source ref', which modifies the working tree and staging area. This is a write operation that changes file contents and index state.
Attacks that exploit this kind of access
Restore paths in worktree and/or index from current state or source ref. It is categorised as a Write tool in the Git MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the Git MCP server in PolicyLayer and add a rule for git_restore: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Git. Nothing to install.
git_restore is a Write tool with medium risk. Write tools should be rate-limited to prevent accidental bulk modifications.
Yes. Add a rate_limit block to the git_restore rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for git_restore. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
git_restore is provided by the Git MCP server (selfagency/git-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →