Skip to content

Why do label-based prompts exclude Linear comments? #1215

@HaiyiMei

Description

@HaiyiMei

Reading the code, I noticed that when a label matches labelPrompts (e.g. Feature → builder), the prompt goes through buildLabelBasedPrompt(), which does not fetch or include Linear comments — while the fallback path (buildIssueContextPrompt()) does.

PR #105 introduced this and described it as "No Linear comments for focused processing." I'd like to understand the reasoning better:

  • What does "focused processing" gain by hiding comments? Was the concern token cost? Distraction / off-topic noise? Something else?
  • Was the assumption that the description is always the source of truth, and comments are sidebar discussion?
  • builder / debugger / scoper system prompts also don't tell Claude to call list_comments itself — so on the label path, Claude has no signal that comments even exist. Is that intentional too?

Asking because in practice we often discuss in comments and then assign (e.g. a blocker gets resolved in a comment, then we hand off to Cyrus). On the label path Cyrus acts on the stale description and misses the resolution. Curious whether this is a use case the maintainers consider in scope, or whether the expectation is "put final scope in the description before assigning."

If the original rationale still holds, no change needed — just want to understand it before working around it. If it's open to revisiting, happy to PR.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions