Continue.dev fails to invoke openspec propose tool with Ollama models (ollama + gemma:e4b)

[tevosouza]

Description

I'm trying to use OpenSpec through Continue.dev with a local Ollama model (gemma4:e4b), but the agent appears to interpret the skill as a terminal command instead of invoking the tool.

As a result, the following error occurs:
open_spec_propose failed because the arguments were invalid, with the following message: Tool open_spec_propose not found. Please try something else or request further instructions.

Environment

• LLM: gemma4:e4b
• Provider: Ollama (local)
• Agent: Continue.dev
• IDE: VSCode

Steps to Reproduce

1. Configure Continue.dev to use Ollama with gemma4:e4b

name: Local Config
version: 1.0.0
schema: v1
models:
  - name: gemma4:e2b
    provider: ollama
    model: gemma4:e2b
    roles:
      - chat
      - edit
      - apply
      - summarize
    capabilities:
      - tool_use

context:
  - provider: repo-map
    params:
      includeSignatures: true
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: codebase

2. Install and configure OpenSpec and run
   openspec init

3. Run the following prompt on agent mode:
   /opsx-propose A microservice called WorkflowService. It should be responsible for creating, managing, and executing generic workflows, such as approval flows like supervisor → coordinator → manager. Use the following technologies: .NET 8, Clean Architecture, CQRS with MediatR, Dapper as the main way to access the database, MassTransit for messaging, FluentValidation and Serilog.

See the magic not happen ;(

I've tested with Qwen3.5, Qwen2.5-Coder (7b / 14b) and Gemma.
I can't get it to work with any of them.

Additional Info

OS: Zorin OS 18.1 x86_64 
Kernel: 6.17.0-20-generic 
Shell: zsh 5.9 
Resolution: 2560x1600, 1920x1080 
DE: GNOME 46.0 
WM: Mutter 
WM Theme: ZorinBlue-Dark 
Theme: ZorinBlue-Dark [GTK2/3] 
Icons: ZorinBlue-Dark [GTK2/3] 
Terminal: vscode 
CPU: 12th Gen Intel i9-12900HX (24) @ 4.900GHz 
GPU: NVIDIA GeForce RTX 4060 Max-Q / Mobile 
GPU: Intel Alder Lake-HX GT1 [UHD Graphics 770] 
Memory: 64034MiB 

[alfred-openspec]

Thanks for flagging this. Continue is a supported surface for us, so this looks worth investigating rather than closing out as model noise. If you can share your Continue version and the generated .continue/prompts/opsx-propose.prompt file, that will help us narrow down whether this is our prompt format or the local-model invocation path in Continue.

[tevosouza]

Hi, thanks for taking a look at this.

The Continue version I’m using is 1.2.22, and I’ve attached the generated .continue/prompts/opsx-propose.prompt file as requested.

Let me know if you need any additional information or if there’s anything else I can test on my side.

opsx-propose.md