Research Review

Research Review via External Reviewer Backend (ultra reasoning)

🔒 Do not wrap this skill in /loop, /schedule, or CronCreate. It is verdict-bearing — it produces a cross-model review verdict, multi-round with reviewer thread continuity. An external timer re-fires the verdict on wall-clock time and breaks the reviewer's round-to-round memory: zero new signal, full token cost. Schedule the external wait that precedes it (work ready → then review once), not the verdict. See shared-references/external-cadence.md.

Get a multi-round critical review of research work from the selected external reviewer backend with maximum reasoning depth.

Constants

• REVIEWER_BACKEND = self — Default when no Codex CLI or other model/API is configured: Claude reviews its own work (no REVIEWER_MODEL needed) at the same ultra-equivalent depth. Override with — reviewer: manual for Manual Review MCP if a genuinely different model is available — that backend uses a model the user chooses, but it must be a recognized model from a different family (OpenAI, Anthropic, Google, DeepSeek, Moonshot/Kimi, Qwen). See auto-review-loop's "Self-Review Backend (No Second Model)" for what self does and the independence tradeoff it carries.

Reviewer Calling Convention

When calling the reviewer, branch on REVIEWER_BACKEND:

If REVIEWER_BACKEND = self (default — no Codex CLI or other model/API configured): No MCP thread is involved. Re-read the briefing fresh from disk (not from memory of writing it) and answer the same prompt a reviewer would get, actively arguing against the work's own claims before accepting them, and flagging any part Claude cannot judge impartially — see auto-review-loop's "Self-Review Backend (No Second Model)" for the full method.

If REVIEWER_BACKEND = manual: Use mcp__manual_review__review for new review threads with: prompt: [the same review prompt used under self above] config: {"model_reasoning_effort": "xhigh", "executor_model": "", "require_reviewer_model": true} Save the returned threadId. Use mcp__manual_review__review_reply for follow-up rounds with: threadId: [saved manual-review threadId] prompt: [follow-up prompt] config: {"model_reasoning_effort": "xhigh", "executor_model": "", "require_reviewer_model": true}

Content fidelity: the manual reviewer should see the same substantive review brief a self-backend read would use. If the manual UI supports file upload / attachment, reuse the same brief file; otherwise paste the brief contents inline because remote web UIs cannot read your local filesystem paths. Review tracing applies to every backend (see Review Tracing below).

Context: $ARGUMENTS

Prerequisites

• None, by default. With no Codex CLI or other model/API configured, REVIEWER_BACKEND = self needs nothing beyond Claude Code itself.
• Optional — Manual Review MCP, for a genuinely independent review when another model is available. Configure per its own setup docs; then pass — reviewer: manual.

Workflow

Step 1: Gather Research Context

Before calling the external reviewer, compile a comprehensive briefing:

1. Read project narrative documents (e.g., STORY.md, README.md, paper drafts)
2. Read any memory/notes files for key findings and experiment history
3. Identify: core claims, methodology, key results, known weaknesses

Step 2: Initial Review (Round 1)

Send a detailed prompt with ultra-equivalent reasoning depth, using the selected backend. Write the full briefing to RESEARCH_REVIEW_REQUEST.md first either way — it's the durable artifact self-review re-reads fresh.

For self backend (default): re-read RESEARCH_REVIEW_REQUEST.md from disk (not from memory of writing it) and answer the prompt below directly — see auto-review-loop's "Self-Review Backend" for the method:

Read the review brief at <absolute path to RESEARCH_REVIEW_REQUEST.md>.
Notes made while producing the work are not evidence beyond the files they
cite, so verify the referenced artifacts before judging.
Please act as a senior ML reviewer (NeurIPS/ICML level). Start from the
assumption that the work is broken somewhere — your job is to find where.
Be adversarial. Trust nothing the work's own narrative tells you — verify
everything yourself. Identify:
1. Logical gaps or unjustified claims
2. Missing experiments that would strengthen the story
3. Narrative weaknesses
4. Whether the contribution is sufficient for a top venue
Please be brutally honest. Flag any point Claude cannot judge impartially
because it produced the work being reviewed.

The review brief should contain the full research context, the specific questions, and the primary artifact / raw-result paths the reviewer should inspect.

For manual backend: use mcp__manual_review__review with the same brief contents. If the manual-review UI supports attachments, attach RESEARCH_REVIEW_REQUEST.md; otherwise paste the brief inline. Save the returned threadId.

Step 3: Iterative Dialogue (Rounds 2-N)

For self backend (default): write an updated brief such as RESEARCH_REVIEW_ROUND_2.md, then re-read it fresh from disk and answer directly — no threadId, so explicitly re-state the prior round's unresolved weaknesses in the prompt rather than relying on thread memory:

Read the updated review brief at <absolute path to
RESEARCH_REVIEW_ROUND_2.md>.
Focus on unresolved weaknesses from the prior round and whether the revision
actually fixed them, not just reworded around them.

For manual backend: use mcp__manual_review__review_reply with the same threadId. Attach that same updated brief if possible; otherwise paste it inline.

For each round:

1. Respond to criticisms with evidence/counterarguments
2. Ask targeted follow-ups on the most actionable points
3. Request specific deliverables: experiment designs, paper outlines, claims matrices

Key follow-up patterns:

• "If we reframe X as Y, does that change your assessment?"
• "What's the minimum experiment to satisfy concern Z?"
• "Please design the minimal additional experiment package (highest acceptance lift per GPU week)"
• "Please write a mock NeurIPS/ICML review with scores"
• "Give me a results-to-claims matrix for possible experimental outcomes"

Step 4: Convergence

Stop iterating when:

• Both sides agree on the core claims and their evidence requirements
• A concrete experiment plan is established
• The narrative structure is settled

Step 5: Document Everything

Save the full interaction and conclusions to a review document in the project root:

• Round-by-round summary of criticisms and responses
• Final consensus on claims, narrative, and experiments
• Claims matrix (what claims are allowed under each possible outcome)
• Prioritized TODO list with estimated compute costs
• Paper outline if discussed

Update project memory/notes with key review conclusions.

Composed mode — if invoked with — composed: <canonical-report-path> (an orchestrator like /idea-discovery passes this), do not write a standalone review .md in the project root. The raw conversation is already persisted to .aris/traces/… (see Review Tracing below — that audit copy is kept in every mode); fold the review conclusions (consensus, claims matrix, prioritized TODOs) into the orchestrator's canonical report and cite the trace path there. Default (no — composed: directive): behave exactly as above — write the standalone review document. Never infer composed mode from a report file merely existing. Full rules: shared-references/output-composition.md.

Key Rules

• Under self (default), review at the same depth every round implies: re-read fresh, argue against the work, flag impartiality gaps — see auto-review-loop's "Self-Review Backend". It is a real check but not an independent one; prefer manual with a genuinely different model whenever one is available.
• For manual, use the identity-bearing config from the Reviewer Calling Convention above (model: gpt-5.6-sol + ultra/xhigh reasoning is the Codex-specific pin from before this conversion and no longer applies under self)
• Put comprehensive context in the review brief either way — self-review reads local files directly; manual reviewers usually cannot, so attach or paste the same brief there.
• Be honest about weaknesses — hiding them leads to worse feedback
• Push back on criticisms you disagree with, but accept valid ones
• Focus on ACTIONABLE feedback — "what experiment would fix this?"
• Document the threadId for potential future resumption (manual backend only)
• The review document should be self-contained (readable without the conversation)

Prompt Templates

For initial review:

"I'm going to present a complete ML research project for your critical review. Please act as a senior ML reviewer (NeurIPS/ICML level)..."

For experiment design:

"Please design the minimal additional experiment package that gives the highest acceptance lift per GPU week. Our compute: [describe]. Be very specific about configurations."

For paper structure:

"Please turn this into a concrete paper outline with section-by-section claims and figure plan."

For claims matrix:

"Please give me a results-to-claims matrix: what claim is allowed under each possible outcome of experiments X and Y?"

For mock review:

"Please write a mock NeurIPS review with: Summary, Strengths, Weaknesses, Questions for Authors, Score, Confidence, and What Would Move Toward Accept."

Review Tracing

A self round has no MCP thread to trace — instead record reviewer_backend: self and independence_verified: false alongside the round's conclusions in the review document. After each mcp__manual_review__review or mcp__manual_review__review_reply call, save the trace following shared-references/review-tracing.md (Policy C — forensic; never silently skip). Use save_trace.sh (resolved per the chain in shared-references/integration-contract.md §2) or write files directly to .aris/traces/<skill>/<date>_run<NN>/. Respect the --- trace: parameter (default: full). A verdict-bearing manual response MUST begin with Reviewer-Model: <exact-model-id> — pass the model THIS session is actually running as in executor_model. Missing, unknown, or same-family identity cannot acquit; emit REVIEW_UNAVAILABLE rather than guessing. If the executor model cannot be named, manual review's cross-family claim is unprovable — say so in the report instead of asserting it.