AI agents call git_diff to retrieve information from Context without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
git_diff retrieves and displays version control metadata about file modifications without altering any state. This is a pure read operation with minimal risk; the output is informational only. Even in a multi-tool agent context, misuse of this tool (e.g., querying unexpected paths) poses no destructive or executable threat.
From the tool's definition Tool description states 'Show file changes' and 'scope to a path' — these are query/retrieval operations. The term 'staged:true for cached diff' refers to examining git's staging area, not modifying it.
Attacks that exploit this kind of access
Show file changes. Use staged:true for cached diff. Optionally scope to a path. It is categorised as a Read tool in the Context MCP Server, which means it retrieves data without modifying state.
Register the Context MCP server in PolicyLayer and add a rule for git_diff: 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 Context. Nothing to install.
git_diff 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_diff 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_diff. 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_diff is provided by the Context MCP server (vibhasdutta/context-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 →