Problem
azure.ai.skills checks the active azd environment for AZURE_AI_PROJECT_ENDPOINT, while current Foundry project provisioning and agent initialization persist the canonical FOUNDRY_PROJECT_ENDPOINT value.
As a result, skill commands and azure.ai.skill service deployment can fail with:
no Foundry project endpoint resolved
even though other Foundry extensions resolve the current project successfully.
Expected Behavior
Skill operations should resolve the project endpoint in this order:
- Explicit
--project-endpoint
- Active azd
FOUNDRY_PROJECT_ENDPOINT
- Active azd
AZURE_AI_PROJECT_ENDPOINT compatibility fallback
- Persisted project context
- Process
FOUNDRY_PROJECT_ENDPOINT
- Process
AZURE_AI_PROJECT_ENDPOINT compatibility fallback
Acceptance Criteria
- Skill commands work when only active azd
FOUNDRY_PROJECT_ENDPOINT is configured.
FOUNDRY_PROJECT_ENDPOINT wins when both active environment keys are set.
- Existing projects that contain only
AZURE_AI_PROJECT_ENDPOINT continue to work.
- The same canonical-first behavior applies to process environment variables.
- Resolver documentation and tests reflect the supported precedence.
Problem
azure.ai.skillschecks the active azd environment forAZURE_AI_PROJECT_ENDPOINT, while current Foundry project provisioning and agent initialization persist the canonicalFOUNDRY_PROJECT_ENDPOINTvalue.As a result, skill commands and
azure.ai.skillservice deployment can fail with:even though other Foundry extensions resolve the current project successfully.
Expected Behavior
Skill operations should resolve the project endpoint in this order:
--project-endpointFOUNDRY_PROJECT_ENDPOINTAZURE_AI_PROJECT_ENDPOINTcompatibility fallbackFOUNDRY_PROJECT_ENDPOINTAZURE_AI_PROJECT_ENDPOINTcompatibility fallbackAcceptance Criteria
FOUNDRY_PROJECT_ENDPOINTis configured.FOUNDRY_PROJECT_ENDPOINTwins when both active environment keys are set.AZURE_AI_PROJECT_ENDPOINTcontinue to work.