You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A fresh audit after the recent fixes found that the project is much healthier overall: build, typecheck, and hook tests are green, and earlier runtime issues around mode detection, data-dir consistency, and standalone behavior appear to be in much better shape.
However, the audit also found a small set of remaining gaps that matter because they affect visible behavior rather than internal plumbing.
These fall into three groups:
Tiny Actor terminal output is partially implemented but still not fully width-safe or fully wired into real Claude Code terminal surfaces.
The new council preset / consensus work exists in isolation but is not yet integrated into the main parse_mode response path.
The plugin build still regenerates a README that drifts away from the current standalone-aware product behavior.
This parent issue tracks those remaining gaps.
Scope
In scope:
Claude Code terminal-rendered plugin output, not the separate TUI
Tiny Actor width safety, runtime integration, and persona fidelity
MCP response integration for council preset / consensus data
Build-generated plugin documentation drift
Out of scope:
The command namespace migration work already tracked elsewhere
Replacing every existing Buddy surface in one pass
Large visual redesign beyond the Tiny Actor rollout
Why this matters
The remaining problems are subtle because many tests already pass. The issue is not that the codebase is unstable; it is that some newly added features are still incomplete at the final integration layer.
That means users can still run into:
crooked Tiny Actor layouts with multilingual text
a feature-flagged preview path that does not show up in normal Claude Code terminal output
generic actor faces instead of the actual agent personas already defined in the repo
council building blocks that exist in code but do not affect the real parse_mode response yet
a build step that rewrites docs back to an older MCP-only stance
Acceptance criteria
This parent issue is complete when all of the following are true:
Tiny Actor terminal output remains aligned under CJK, emoji, and ANSI cases
Tiny Actor rendering uses the width-safe card/grid path instead of ad-hoc preview formatting
Tiny Actor output is reachable from actual Claude Code terminal plugin surfaces behind an explicit flag or rollout gate
Tiny Actor cards use real agent visual.eye metadata where available
council preset / consensus data is emitted from the main MCP workflow in a way clients can actually consume
the generated plugin README no longer reintroduces outdated MCP-required messaging
Background
A fresh audit after the recent fixes found that the project is much healthier overall: build, typecheck, and hook tests are green, and earlier runtime issues around mode detection, data-dir consistency, and standalone behavior appear to be in much better shape.
However, the audit also found a small set of remaining gaps that matter because they affect visible behavior rather than internal plumbing.
These fall into three groups:
parse_moderesponse path.This parent issue tracks those remaining gaps.
Scope
In scope:
Out of scope:
Why this matters
The remaining problems are subtle because many tests already pass. The issue is not that the codebase is unstable; it is that some newly added features are still incomplete at the final integration layer.
That means users can still run into:
parse_moderesponse yetAcceptance criteria
This parent issue is complete when all of the following are true:
visual.eyemetadata where availableSub-issues
visual.eyemetadataparse_moderesponse pathRecommended order