AI agents use commit to create or update resources in Jj — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Jj environment.
The commit operation modifies the repository by recording changes with a message and creating a new change, which are core Write operations in version control. While commits are important checkpoints, they are not destructive (can be amended/abandoned in jj) and not irreversible in the traditional sense.
From the tool's definition Tool description states it 'Update[s] the current change with the specified message' and 'create[s] and move[s] to a new, empty change' in a Jujutsu repository. These are reversible modifications to version control state.
Attacks that exploit this kind of access
Update the current change with the specified message in a Jujutsu (jj) repository and then create and move to a new, empty change. This set of actions is jj. It is categorised as a Write tool in the Jj MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the Jj MCP server in PolicyLayer and add a rule for commit: 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 Jj. Nothing to install.
commit 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 commit 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 commit. 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.
commit is provided by the Jj MCP server (jj-mcp-server). 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 →