viktor/beadboard
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
beadboard/docs/prompts/2026-02-28-next-session-holistic-ux-critique.md

4.3 KiB
Raw Blame History

Next Session Prompt: Holistic UX Critique for Professional Multi-Agent Operations

You are continuing work in beadboard.

Understanding Brief

BeadBoard is intended to be a professional multi-agent communication + work management system where:

  • agents coordinate through Beads (bd) state,
  • humans supervise, intervene, and steer,
  • both can collaborate across multiple repos/projects.

This is not just a task board. It is an operations surface for swarm-style execution, communication, assignment, and recovery.

Recent work completed (must understand before critique)

  1. bd/Dolt recovery was performed after severe divergence:
    • prior broken state: Dolt showed only a few issues while historical stores had hundreds,
    • repaired state now shows healthy inventory (381 issues) and non-empty bd ready.
  2. Runtime assumptions changed:
    • bd upgraded to 0.56.1.
    • Dolt server mode is operationally relevant (127.0.0.1:3307).
  3. bd command execution was moved toward PATH-based portability and clearer setup failure handling.

Do not re-do this recovery unless evidence shows a new regression.

First required step (before UX critique)

Audit current local uncommitted work and summarize it in UX terms.

Run and analyze at minimum:

  • git status --short
  • git diff --stat
  • targeted diffs for UX-facing areas (views, tabs, drawers, assignment, session feed, top/nav shell)

Then produce:

  1. what changes are incomplete but directionally correct,
  2. what changes conflict with intended product behavior,
  3. what changes increase accidental complexity.

Product intent to evaluate against

Evaluate all views/tabs as one cohesive system for:

  • Agent-first operations (fast, low-friction, low ambiguity)
  • Human oversight (clear state, intervention points, confidence)
  • Cross-project/scoped execution (project scope switching without confusion)
  • Communication reliability (comments/messages/coordination context where decisions happen)

Critique targets (must cover all)

  1. Information architecture across main views/tabs.
  2. Distinct role of each major surface (Social, Graph, Sessions, side panels, drawers).
  3. Assignment UX consistency and discoverability.
  4. Communication model UX (threads/comments/agent interactions) and where it breaks flow.
  5. State clarity: ready vs blocked vs in-progress; ownership; handoff visibility.
  6. Failure mode UX (server unavailable, path/config mismatches, stale data indicators).
  7. Cognitive load: where operators need to context-switch too much.
  8. Terminology consistency (bead, task, swarm, molecule, session, agent).

Required outputs

1) UX Critique Report

Provide a structured critique with:

  • What is working (keep)
  • What is ambiguous (clarify)
  • What is broken/risky (fix)
  • severity per issue (P0, P1, P2)
  • concrete file/component references

2) Target UX Model (recommended)

Propose a clean target model for view responsibilities:

  • one-line purpose per view/tab,
  • key interactions per surface,
  • interactions that must be shared vs isolated.

3) Prioritized execution backlog

Create beads for follow-up work from critique findings:

  • one bead per coherent unit,
  • include scope/out-of-scope and acceptance criteria,
  • preserve dependency correctness.

4) Minimal change strategy

Recommend a staged rollout plan that avoids large regressions:

  • phase 1: low-risk high-value consistency fixes,
  • phase 2: IA/view role cleanup,
  • phase 3: deeper workflow refinements.

Constraints

  • Preserve current route model (/ with view= query params).
  • Keep changes grounded in actual implemented code (no speculative claims).
  • Reuse shared components/logic; avoid one-off behavior per view.
  • Keep language simple and operator-facing.
  • Approval gate: Do not create any beads during discovery/brainstorming. First present findings + proposed bead backlog draft, then wait for explicit user approval before running any bd create commands.

Quality bar

The critique should read like a professional product/UX architecture review for an agent operations platform, not generic UI feedback.

Completion criteria

  • Clear diagnosis of current UX shape using actual uncommitted code.
  • Decision-ready target model for views/tabs and communication surfaces.
  • Prioritized, execution-ready bead backlog generated from findings.