microsoft / microsoft/GitHub-Copilot-for-Azure
[Skill Bug]: cost related prompts are routed to skills other than azure-cost
- Dominant language
- Python
- Stars
- 250
- Forks
- 204
- Avg merge
- 1d 12h
- Merged PRs (30d)
- 67
Description
### Preflight Checklist
- [x] I have installed the [latest version of GitHub Copilot for Azure in VS Code](https://github.com/microsoft/GitHub-Copilot-for-Azure/releases/latest).
- [x] I have checked existing resources, including using the /help command in the chat window.
- [x] I have searched for [similar issues](https://github.com/microsoft/GitHub-Copilot-for-Azure/issues).
### Repro Steps
This is a significant behavior change after upgrading from Copilot SDK from 0.3.x to 1.0.x.
There have been two patterns:
- Skill routed to azure-diagnostics
- Skill routed to service specific skill that relates to the service in the prompt
See test runs for azure-costs starting from 06/22/2026.
Example runs:
- https://agreeable-dune-0f718070f.1.azurestaticapps.net/nightly-runs.html?file=2026-06-20%2F27871296299%2Fazure-cost%2Fazure-cost_Storage_lifecycle_policy_optimization%2Fagent-metadata-2026-06-20T12-49-53-182Z.md
- https://agreeable-dune-0f718070f.1.azurestaticapps.net/nightly-runs.html?file=2026-06-20%2F27871296299%2Fazure-cost%2Fazure-cost_AKS_cost_spike_investigation%2Fagent-metadata-2026-06-20T12-45-21-412Z.md
### Prompt
See test runs
### Expected Result
We should either fix the routing, or make sure each skill has reference path to present the same content or point back to azure-costs skill.
### Actual Result
See above
### Attachments
_No response_
### Regression?
behavior change after upgrading from Copilot SDK from 0.3.x to 1.0.x.
### Additional Context
_No response_
Contributor guide
Research direction
Start by comparing the linked nightly test runs and the routing behavior after the Copilot SDK upgrade from 0.3.x to 1.0.x. Trace how cost-related prompts are selected between azure-cost, azure-diagnostics, and service-specific skills; done means those prompts route to azure-cost or consistently expose equivalent content or a link back to it.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- azure, python
- Domain
- ai, devtools
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100