AI agents call git_reflog to retrieve information from Git without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
git_reflog retrieves and displays historical information about HEAD and ref movements. It performs no writes, deletions, or code execution. While it may reveal sensitive commit history or branching patterns, its core function is purely informational and safe for agent use. The safety risk is minimal compared to tools like git_commit, git_push, or git_reset which modify state.
From the tool's definition Tool description states 'Show local HEAD and ref movement history' — a query/viewing operation with no modification capability. The reflog is a read-only inspection of Git's internal history log.
Attacks that exploit this kind of access
Show local HEAD and ref movement history for recovery. It is categorised as a Read tool in the Git MCP Server, which means it retrieves data without modifying state.
Register the Git MCP server in PolicyLayer and add a rule for git_reflog: 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_reflog is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the git_reflog 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_reflog. 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_reflog 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 →