⚠️ No Changeset found

Latest commit: 9d4ad13

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

@snoyiatk thanks for the PR? Does it change anything in your eyes now that we support rules folders, or do you still want more flexibility to point at specific files?

@mrubens I justed found that new PR #2354 few hours ago, to me it partially resolve (1), so this PR might still help with (2).
What do you think?

Also, #2354 currently prioritize .roo/rules-{mode} over .roo/rules, which I think should be the opposite? (.roo/rules is common and should be loaded first before .roo/rules-{mode}
And there is bug mentioned in the PR.
Will submit another PR for that.

Today I tried to refactor RooFlow custom instructions follow our new .roo/rules and .roo/rules-{mode} structure, and as expected, there is still many duplicate instructions no matter what combination I tried.
For example
Memory-bank instructions: Some mode allow read to memory-bank, some doesnt.
Without custom path configuration, we end up placing same memory-bank instructions in at least 3 rules folder like rules-{code/debug/test}, different memory-bank instruction for ask/architecture
Same thing for mcp/tools, for most of mode I want to allow mcp/tools, but for some-mode I don't want it.

There are other cases that I found default folder structure couldn't fit and I want to customize the paths as my need.

Hi @snoyiatk, I just merged a PR to support symbolic links in rules folders/files. Does that help at all? #2421

@mrubens yeah that was one of the solution I considered but ended up didn't choose it.
Will have a try to see if it's good 👌

Hi! I'm new to Roo and experimenting with its configuration.
I’m wondering if this feature can be used in the following way:

    {
      "customInstructions": "No rules",
      "customInstructionsPaths": [
        "~/.roo/ask-no-code/system-prompt-ask-no-code"
      ],
      "groups": [
        "read"
      ],
      "name": "Ask(No System)",
      "roleDefinition": "You are Roo, an experienced technical leader who is inquisitive and an excellent planner. Your goal is to gather information and get context to create a detailed plan for accomplishing the user's task, which the user will review and approve before they switch into another mode to implement the solution.",
      "slug": "ask-no-code",
      "source": "global"
    }

I’m using the global ~/.roo directory for this setup. My question is: will this configuration override the current system prompt?

Currently, it seems the only way to replace the system prompt is by adding a .roo/system-prompt-{slug} file in each workspace. However, I have multiple workspaces scattered across different locations, and managing this in every workspace feels inefficient. Would using the global ~/.roo directory with customInstructionsPaths allow me to set a custom system prompt once and apply it across all my workspaces?"

I have a suggestion to omit default roo code system prompt for mode like this:

    {
      "customInstructions": "No rules",
      "defaultSystemPrompt": false,
      "customInstructionsPaths": [
        "~/.roo/00-common/00-system-info",
        "~/.roo/30-test/system-prompt-test"
      ],
      "groups": [
        "read"
      ],
      "name": "Ask(No System)",
      "roleDefinition": "You are Roo, an experienced technical leader who is inquisitive and an excellent planner. Your goal is to gather information and get context to create a detailed plan for accomplishing the user's task, which the user will review and approve before they switch into another mode to implement the solution.",
      "slug": "ask-no-code",
      "source": "global"
    }

This way we can add back customize system prompt by ourself

Thank you anyway!

Edit: don't do this

@fatihaziz you can globally override prompts for the built-in modes by editing global modes and placing a mode in with the same slug:

    {
      "slug": "architect",
      "name": "Architect",
      "roleDefinition": "You are Roo, an experienced technical leader who is inquisitive and an excellent planner. Your goal is to gather information and get context to create a detailed plan for accomplishing the user's task, which the user will review and approve before you to implement the solution.",
      "customInstructions": "## **Architect Role: Adaptive Planning & Coordination**\n\n**Research & Planning (Adaptive)**\n- Gather context. Perform deeper research and evaluate approaches for complex tasks. Recommend a solution. You have access to product and dependency documentation with the External Documentation MCP server and you can fetch any documentation you might need for future usage. Search documentation often and don't hestiate to search multiple indices. Always search documentation for API calls.\n- Iterate with the user (Strawgate) on complex/ambiguous tasks.\n- **Lite-Touch:** For simple tasks, create a concise `/plans/<plan name>/plan.md`. Execute simple steps directly if appropriate.\n- **Heavy-Touch:** For complex tasks needing delegation, create a detailed `plan.md` and proceed to Implementation Planning.\n\n**Implementation Planning (Heavy-Touch Only)**\n- Define required worker resources (e.g., `Worker 1: Code Expert + Python`). Be practical.\n- Create worker plans (`plan-1.md`, `plan-2.md` etc.) and a verification plan (`plan-verify.md`).\n\n**Context Handoff (If Necessary)**\n- Monitor context size. If approaching limits (~100k tokens), propose handoff to the user.\n- If approved, create `/plans/<plan name>/handoff-<timestamp>.md` with standard fields (Plan Link, Status, Remaining, Context, Next Step) and notify the Project Manager to initiate the next task segment. Cease work on this segment.",
      "groups": [
        "read",
        "edit",
        "browser",
        "command",
        "mcp"
      ],
      "source": "global"
    }

@fatihaziz I lied, don't do this, it updates the UI but the agent is not passed the instructions which is very sad and a bug

@fatihaziz I lied, don't do this, it updates the UI but the agent is not passed the instructions which is very sad and a bug

ohh ok haha,,
maybe we should open a different issue for that?

i have tried it, when I check the preview prompt. the solition is not replacing built in instruction. instead it adds as additional prompt

Hey, @snoyiatk
Did you get a chance to try the new symbolic link feature from #2421?

Hey @snoyiatk,
Thank you for your contribution, We noticed this PR is stale and will be closed. If you plan to revisit this, please create an issue first as required by our issue-first approach before opening a new PR.