Copilot can now name your internal agent tools.
Custom skills and MCP servers no longer have to disappear into an “other” bucket. Usage trends become easier to interpret, with limits.
What changed
GitHub’s usage reports now show configured names for custom skills, agents, MCP servers, commands and plugins. Coverage depends on the client and version; earlier processed days retain the old grouping. GitHub warns that telemetry can undercount.
Go to the source ↓Why it matters
Named activity can answer a useful operational question: are people finding the internal tool a team built? A persistent generic bucket makes that question harder to investigate.
Treat the report as a map of adoption. A low count might mean poor discoverability, an outdated client or a tool that is rarely needed. A high count still says little about whether the work was correct.
What remains unproven
These are client usage signals, not exact accounting or a productivity evaluation. Reports require the relevant policy and permissions, and built-in IDE chat telemetry does not provide the same name coverage.
The useful next step
Use consistent, non-sensitive tool names. Check supported clients and telemetry settings before comparing periods. Pair usage with a review of completed tasks; do not rank people or tool quality by interaction count.
Use named telemetry to find questions, then inspect the work.