phase-10h-chat-integrity-review — Claude Code prompt
phase-10h-chat-integrity-review reviews uncommitted git diffs for Phantom chat transcript integrity in Claude Code. It inspects attachment persistence and message rendering without modifying project code.
Apache-2.0 ★ 1,497 updated 2026-06-16
.claude/commands/phase-10h-chat-integrity-review.mdWhat phase-10h-chat-integrity-review does
Turns Claude Code into a strict technical reviewer for uncommitted changes in git diff. The agent reviews Phase 10H documentation first. Then it checks chat transcripts and grades findings from P0 to P3.
Application files remain untouched because the prompt forbids direct edits. All findings go into phase-10h-chat-integrity-review.md. The instructions target silent file upload failures, unauthenticated attachment previews, and TypeScript escapes like any or @ts-ignore.
The prompt references hardcoded local paths like /Users/truffle/work/phantom-murph-hardening. Running it elsewhere requires updating those paths manually.
What it helps with
- audit uncommitted chat diffs for streaming and replay regressions
- catch silent file upload failures and double-submission issues
- verify attachment metadata persistence without raw base64 payloads
- flag prohibited TypeScript escapes like
anyand@ts-ignore
Example prompts for the agent
- Review the current
git diffusingphase-10h-chat-integrity-reviewand list all regressions. - Run the chat integrity review and write findings to
phase-10h-chat-integrity-review.mdwithout modifying source files.
How to install phase-10h-chat-integrity-review
Manually: put the files in place
.claude/commands/phase-10h-chat-integrity-review.mdHow to use
Paste the text at the start of a chat or into project instructions
Questions about phase-10h-chat-integrity-review
How to install phase-10h-chat-integrity-review in Claude Code?
Save the prompt file to .claude/commands/phase-10h-chat-integrity-review.md. You can then call the command directly in Claude Code.
What does phase-10h-chat-integrity-review do?
It instructs the assistant to inspect your uncommitted git diff, verify chat transcript safeguards, and document issues by severity from P0 to P3.
Does this prompt modify source code?
No, the prompt explicitly instructs the assistant not to touch application files. Its sole output is an evaluation report saved as a Markdown document.
Source text
phase-10h-chat-integrity-review.md · Apache-2.0
ultrathink. ultrathink. ultrathink. You are a principal engineer, principal architect, and principal product manager at Anthropic, all three at once. You are reviewing the Phase 10H Phantom chat transcript-integrity slice in `/Users/truffle/work/phantom-murph-hardening`. 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. ## Mission Review the current uncommitted diff. Do not edit application code. Write your review report to: `/Users/truffle/work/phantom-murph-hardening/research/chat-experience/phase-10h-chat-integrity-review.md` Focus on bugs, regressions, missing tests, and product-trust issues. ## Required Reading Read: 1. `/Users/truffle/.claude/AGENTS.md` 2. `/Users/truffle/.claude/CLAUDE.md` 3. `/Users/truffle/work/phantom-murph-hardening/CLAUDE.md` 4. `/Users/truffle/work/phantom-murph-hardening/research/chat-experience/phase-10h-product-direction.md` 5. `/Users/truffle/work/phantom-murph-hardening/research/chat-experience/phase-10h-phantom-chat-review.md` 6. `/Users/truffle/work/phantom-murph-hardening/research/chat-experience/phase-10h-pi-thinking-research.md` 7. `/Users/truffle/work/phantom-murph-hardening/research/chat-experience/phase-10h-provider-thinking-research.md` 8. `/Users/truffle/work/phantom-murph-hardening/prompts/phase-10h-chat-integrity-builder.md` Then inspect the actual diff with `git diff`, not summaries. ## Review Scope Pay special attention to: - Runtime or SDK result errors must not become empty successful assistant rows. - User attachment metadata must be persisted without base64 payloads. - Sent user-message attachments must appear live and after reload. - Upload failures must not silently send prompts without the intended files. - Assistant messages with multiple text blocks must render all text. - Chat input should not double-send while upload is pending. - Attachment preview URLs must stay authenticated and safe. - Existing chat streaming, replay, and resume semantics must not regress. - No explicit `any`, `@ts-ignore`, hidden type escapes, or broad unrelated refactors. - No generated build artifact should be included unless there is a clear reason. ## Output Format Findings first, ordered by severity. Use P0/P1/P2/P3. Include exact file paths and line references. If there are no P0/P1/P2 findings, say that explicitly. Then include: 1. Tests you ran or inspected. 2. Residual risks. 3. Whether this diff is safe to proceed to live browser verification. Do not make code changes. Write only the report file. ultrathink. Treat the transcript as a trust surface. The review should be skeptical, specific, and grounded in actual files.
© ghostwright · Apache-2.0 · ghostwright/phantom