Govern tool execution

Outcome: demonstrate all permission/approval outcomes and prove only the approved mutation executes.

Module: examples.governed_tool_agent

python examples/governed_tool_agent.py

Expected terminal result includes allow=completed, deny=permission_denied, approve=completed, reject=completed, expiry=approval_expired, and mutations=1.

Muru Workspace shows requested tools, the pending approval card, actor decisions, expiry, and terminal run state. Rejection is a reviewed tool outcome and the agent may continue; expiry fails the run. Missing grants fail before invocation. Redacted arguments, not raw sensitive inputs, enter public events.

Qualified by tests/qualification/test_scenarios.py::test_governed_tool_scenario_exercises_every_policy_outcome and the Runtime approval/tool suites.