Skip to main content
prompt GP ChatGPT

phase-10a-chat-pipeline-research — Claude Code prompt: pipeline audit

phase-10a-chat-pipeline-research — a slash command that walks Phantom's chat pipeline from SDK events through SSE to React rendering. You get a file-grounded report, and no code is touched.

Apache-2.0 ★ 1,497 updated 2026-06-16

Install into Claude Code · Manually: put the files in place All methods and platforms ↓
.claude/commands/phase-10a-chat-pipeline-research.md
Open on GitHub

What phase-10a-chat-pipeline-research does

The prompt puts the model in three roles at once: principal engineer, architect and product manager. It then points it at Phantom's web chat stack, from SDK event ingestion through SSE and reducer state down to React rendering. Token limits, time pressure and any thought of a v2 are ruled out up front. Source files stay untouched; the deliverable is findings.

A reading order is part of the job: 19 files, opening with CLAUDE.md and README.md and ending at chat-store.ts, tool-call-card.tsx and e2e/chat.spec.ts. Nearby files are located with rg. The report must cover a current-state pipeline map, a severity-ordered bug list, a first build slice with tests, a second-stage architecture and explicit non-goals. Notes belong under research/chat-experience/.

Every path is absolute, /Users/truffle/work/phantom-chat-experience/... — rewrite them for your own checkout before running it, or the agent spends the turn opening files that aren't there.

What it helps with

  • Trace a message from the SDK event to the React component
  • Spot wire frames the frontend never renders
  • Explain why the chat looks idle during a long turn
  • List the invariants that must survive a redesign
  • Scope a first build slice with tests

Example prompts for the agent

  • /phase-10a-chat-pipeline-research audit the chat pipeline and give me findings with file references
  • Work out why the UI goes quiet on a long Murph turn, read only, no edits
  • /phase-10a-chat-pipeline-research map the event path from SDK to React and rank the risks

How to install phase-10a-chat-pipeline-research

Manually: put the files in place

.claude/commands/phase-10a-chat-pipeline-research.md

Questions about phase-10a-chat-pipeline-research

How do I install phase-10a-chat-pipeline-research in Claude Code?

Place the file at .claude/commands/phase-10a-chat-pipeline-research.md inside your project. It then runs as /phase-10a-chat-pipeline-research.

What does phase-10a-chat-pipeline-research do?

It runs a read-only audit of Phantom's chat pipeline over a fixed list of 19 files and returns an event map, ranked risks and a first implementation slice with tests. Nothing in the source tree is edited.

Can I run it against a different repository?

The reading list uses absolute paths under /Users/truffle/work/phantom-chat-experience/, so swap them for your own checkout first. Otherwise the agent looks for files that don't exist on your machine.

Source text

phase-10a-chat-pipeline-research.md · Apache-2.0
ultrathink. ultrathink. ultrathink.

You are a principal engineer, principal architect, and principal product
manager at Anthropic, all three at once. Your mission is to audit Phantom's
current web chat experience pipeline from backend SDK event ingestion through
SSE, frontend reducer state, and React rendering. Treat this as a production
agent-chat product, not a cosmetic chat bubble exercise.

No context anxiety. No token limits. No time pressure. No cost anxiety. No
"v2" thinking. There is no v2. Build it right. Take as long as you need.

## Context

Cheema and the orchestrator have just proven Phantom running on Murph with
OpenAI `gpt-5.5` in a real browser. The runtime is working, including
auto-compaction. The user-visible weakness is that long-running tool-heavy
tasks can look dead while Murph and Phantom are actively working.

The goal of this research slice is to understand the existing Phantom chat
pipeline deeply enough to design a best-in-class activity surface. Do not
implement code. Do not edit source files. Produce findings only.

## Required Reading

Read these files in this order:

1. `/Users/truffle/work/phantom-chat-experience/CLAUDE.md`
2. `/Users/truffle/work/phantom-chat-experience/README.md`
3. `/Users/truffle/work/phantom-chat-experience/src/chat/types.ts`
4. `/Users/truffle/work/phantom-chat-experience/src/chat/types-tool.ts`
5. `/Users/truffle/work/phantom-chat-experience/src/chat/sdk-to-wire.ts`
6. `/Users/truffle/work/phantom-chat-experience/src/chat/sdk-to-wire-handlers.ts`
7. `/Users/truffle/work/phantom-chat-experience/src/chat/writer.ts`
8. `/Users/truffle/work/phantom-chat-experience/src/chat/http.ts`
9. `/Users/truffle/work/phantom-chat-experience/src/chat/stream-sse.ts`
10. `/Users/truffle/work/phantom-chat-experience/src/chat/event-log.ts`
11. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/hooks/use-chat.ts`
12. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/lib/chat-store.ts`
13. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/lib/chat-dispatch-tools.ts`
14. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/components/message-list.tsx`
15. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/components/assistant-message.tsx`
16. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/components/tool-call-card.tsx`
17. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/components/thinking-block.tsx`
18. `/Users/truffle/work/phantom-chat-experience/chat-ui/src/lib/__tests__/chat-store.test.ts`
19. `/Users/truffle/work/phantom-chat-experience/e2e/chat.spec.ts`

Also inspect any adjacent files that are directly needed to answer the
questions below. Use `rg` first.

## Questions To Answer

1. What is the exact event path from SDK/Murph message to the UI? Include each
   file and function in order.
2. Which backend wire frames exist today but are ignored or under-rendered by
   the frontend?
3. Which frontend state objects disappear at `session.done`, and does that
   erase useful post-run tool history from the transcript?
4. Why can the UI look idle during a long-running Murph turn? Identify every
   contributing cause, not just the first one.
5. What invariants must not break when we redesign this? Include SSE resume,
   persisted messages, assistant lifecycle, tool-call protocol, thinking
   blocks, error rows, and abort behavior.
6. What would a minimal but production-grade first implementation look like?
   Keep it scoped and testable. Name exact files likely to change.
7. What would a larger best-in-class implementation look like after that?
   Separate necessary product quality from speculative polish.

## Acceptance Criteria

Your final answer must be findings-first, with concrete file references and
line-level anchors when possible. It must include:

1. A current-state pipeline map.
2. A bug/risk list ordered by severity.
3. A recommended first build slice with tests.
4. A second-stage product architecture for richer observability.
5. Explicit non-goals for this phase.

Do not include generic chat UX advice. Every claim must be grounded in Phantom
source code you read.

## Anti-Patterns

Do not recommend a rewrite unless you can name the specific contract the
rewrite improves and the migration path that keeps existing behavior working.
Do not add reasoning to TypeScript that belongs in the agent. Do not ask for
server changes unless the current wire protocol cannot carry the needed state.
Do not trust comments without reading the implementation.

## Handoff

Return your report in the final response. If you create notes, put them under
`/Users/truffle/work/phantom-chat-experience/research/chat-experience/`.

ultrathink. Stay product-minded and implementation-grounded.

© ghostwright · Apache-2.0 · ghostwright/phantom

Similar prompts

All prompts for Agent setup: memory, context, orchestration →

From the same set38

The whole set →
echo Before answering a substantive question, echo searches memory for a similar earlier decision. If it finds a strong match, it brings up that conclusion and asks whether it still applies. CX Codex +1 ★ 1,497repo stars phase-10m-pi-murph-continuity-research phase-10m-pi-murph-continuity-research is a prompt for Claude Code and ChatGPT that investigates context recovery and compaction in Murph and Phantom, checking whether they reinvent Pi primitives. GP ChatGPT ★ 1,497repo stars phase-10i-transcript-recovery-builder The phase-10i-transcript-recovery-builder prompt for Claude Code builds transcript recovery for Phantom on Murph, letting agents search chat history after context compaction. GP ChatGPT ★ 1,497repo stars thread Trace how your thinking on a topic changed over time. The skill turns memory entries into a chronological account with marked turning points. CX Codex +1 ★ 1,497repo stars overheard overheard scans two weeks of stored memory to surface unfulfilled commitments and prepares followup drafts. CX Codex +1 ★ 1,497repo stars phase-10d-chat-ui-polish-research The phase-10d-chat-ui-polish-research prompt guides a review of Phantom’s chat UI using screenshots and project files, then asks for a design direction and an implementation plan. GP ChatGPT ★ 1,497repo stars ritual Detects recurring behaviors from your session history and transforms them into scheduled tasks. If you regularly do something on a cadence, ritual proposes automating it. CX Codex +1 ★ 1,497repo stars show-my-tools show-my-tools audits the agent environment and lists loaded skills, memory files, plugins, hooks, and subagents. It also outputs a settings summary and dashboard links. CX Codex +1 ★ 1,497repo stars

Project sponsors

Companies that keep this analytics open