-
Notifications
You must be signed in to change notification settings - Fork 3.9k
Open
Labels
area:toolsRelates to tool usageRelates to tool usage
Description
Before submitting your bug report
- I've tried using the "Ask AI" feature on the Continue docs site to see if the docs have an answer
- I'm not able to find a related conversation on GitHub discussions that reports the same bug
- I'm not able to find an open issue that reports the same bug
- I've seen the troubleshooting guide on the Continue Docs
Relevant environment info
- OS: any
- Continue version: 1.3.28 (verified against main branch)
- IDE version: any
- Model: any (all models receive conflicting instructions)
- config: anyDescription
Response Formatting Instructions Mixed Into Tool Descriptions
Summary
Tool descriptions contain instructions about how to format response text, which conflates "how to use this tool" with "how to write your response."
Example
"When suggesting subsequent shell commands ALWAYS format them in shell command blocks."
This instruction is about formatting the assistant's prose response, not about how to call the run_terminal_command tool.
The Problem
- Separation of concerns: Tool descriptions should describe the tool's purpose and parameters, not response formatting
- Maintenance: Formatting preferences are scattered across tool definitions instead of centralized
- Confusion: Models may interpret this as a tool-calling instruction rather than a formatting preference
Proposed Fix
Move response formatting instructions to the base system message or a dedicated formatting section, keeping tool descriptions focused on:
- What the tool does
- What parameters it accepts
- When to use it vs alternatives
- Any constraints or limitations
Affected Tools
run_terminal_command- "ALWAYS format them in shell command blocks"- Potentially others with similar mixed concerns
To reproduce
No response
Log output
Metadata
Metadata
Assignees
Labels
area:toolsRelates to tool usageRelates to tool usage
Type
Projects
Status
Todo