review
Launch a multi-agent editorial review with 8 specialists (2 literature scouts plus 6 domain reviewers) orchestrated by an area-chair agent. Produces a unified report plus specialist reports in .papermill/reviews/. Updates .papermill/state.md.
- License 许可
- MIT
- 最近更新
- 2026-10-23
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- review
- description
- Launch a multi-agent editorial review with 8 specialists (2 literature scouts plus 6 domain reviewers) orchestrated by an area-chair agent. Produces a unified report plus specialist reports in .papermill/reviews/. Updates .papermill/state.md.
03 SKILL.md 正文
Multi-Agent Editorial Review
Launch a comprehensive multi-agent review of a research paper. The review system uses 8 specialist agents orchestrated by an area chair to evaluate logic, novelty, methodology, prose, citations, and formatting: grounded in literature context.
Step 1: Read Context
Read .papermill/state.md (Read tool) for:
- Thesis: What the paper claims (the review checks if the paper delivers).
- Venue: Target venue (review against its standards).
- Review history: Previous reviews and their findings.
- Format: Paper format (latex, markdown, rmarkdown).
If .papermill/state.md does not exist, the review can still proceed by reading the manuscript directly, but the review will be less targeted without thesis and venue context. Note this limitation to the user and suggest running /papermill:init first for best results.
Step 2: Identify the Manuscript
Locate the manuscript files:
- Check
.papermill/state.mdfor format and any recorded manuscript path. - Scan for manuscript files (Glob tool):
*.tex,*.Rmd,paper.md,manuscript.md. - Identify the main file and any supporting files (bibliography, figures, included files).
Read the manuscript to confirm it has enough content for review.
Step 3: Ask for Focus Areas
Before launching the review, ask the user:
I'll launch a multi-agent review with specialists covering:
- Logic & proofs: mathematical correctness and argument structure
- Novelty: contribution evaluation against the literature
- Methodology: experimental design and statistical rigor
- Prose: writing quality and narrative structure
- Citations: reference accuracy and completeness
- Formatting: build verification and venue compliance
Are there specific areas you want me to focus on, or should I run the full review?
If the user specifies focus areas, note them for the orchestrator. If they want the full review, proceed with all specialists.
Step 4: Launch the Reviewer Orchestrator
Launch the reviewer agent (Agent tool with subagent_type: "papermill:reviewer").
Pass the agent:
- Path to the manuscript file(s)
- Path to
.papermill/state.md(if it exists) - Any user-specified focus areas or tone preferences
- The thesis statement and target venue (if known)
The agent will:
- Read the paper and produce a comprehension summary
- Spawn 2 literature scouts in parallel for field context
- Spawn 6 specialist reviewers in parallel
- Cross-verify critical findings
- Write individual specialist reports and a unified report to
.papermill/reviews/YYYY-MM-DD-HHMM/, with stable finding ids (F1, F2, ...) - Return the
review.mdpath, recommendation, and counts
Step 5: Present Results
After the agent completes, read the review.md path it returned (Read tool).
Present a summary to the user:
Review Complete
Recommendation: [ready | minor-revision | major-revision | not-ready]
Severity Count Critical N Major M Minor P Suggestions Q Top findings:
- [Most important finding]
- [Second most important finding]
- [Third most important finding]
The full report is at
[returned review.md path]. Individual specialist reports are in the same directory.
Then ask: "Would you like to go through the findings in detail, or address the critical issues first?"
Step 6: Update State File
Update .papermill/state.md (Edit tool). If .papermill/state.md does not exist, skip this step and suggest /papermill:init. Updates:
Add a review record to review_history:
review_history:
- date: "YYYY-MM-DD"
type: "multi-agent-review"
findings_critical: K
findings_major: N
findings_minor: M
recommendation: "ready | minor-revision | major-revision | not-ready"
notes: "Brief summary of key findings"
report_path: "[returned review.md path]"
Advance stage to review unless it is already review or submission (never move it backward).
Append a timestamped note to the markdown body.
Step 7: Suggest Next Steps
Based on the recommendation, suggest the most relevant next step:
- Ready for submission →
/papermill:polishto prepare, then/papermill:venueif no venue is selected yet. - Minor revision →
/papermill:reviseto apply the fixes with verification, then re-run/papermill:review. - Major revision →
/papermill:revisefor the concrete findings. If issues are structural, suggest/papermill:outlinefirst; if issues are with the argument, suggest/papermill:thesis. Then re-run/papermill:review. - Not ready → Identify the root cause. Weak thesis →
/papermill:thesis. Insufficient evidence →/papermill:experimentor/papermill:simulation. Missing related work →/papermill:prior-art. Proof gaps →/papermill:proof.
04 同仓库更多 Skills
worldsmith-methodology
Use when working in a project with a worldbuilding documentation ecosystem (docs/ or lore/ directory). Provides the documentation-first methodology: canonical hierarchy, doc roles (timeline, lore, characters, systems), propagation discipline, and the dual canonical and exploratory workflow.
prose-craft
Use when writing fiction, editing fiction prose, reviewing chapters, or drafting scenes. Covers prose craft, dialogue mechanics, scene structure, and common failure modes in AI-assisted fiction. Trigger phrases: "write fiction", "edit chapter", "prose review", "scene draft", "dialogue check", "fiction craft", "edit this scene", "review my prose".
vista
This skill should be used when the user asks about research directions, open problems, future work, or follow-up research from the academic literature. Trigger phrases include "open problems in X", "what's next for Y", "future work for paper Z", "research directions on T", "salient follow-up research", "what should I work on next in", "find me follow-up research unrelated to my prior work", "broad survey of W", "neglected directions in V". Routes the request to the right Vista MCP tool, reads back the structured paper sections, and synthesizes research directions in the conversation.
synthesize
Make a research-agent run conclude and write synthesis.md (outcome, key findings, failed approaches, open questions, recommendations). Signals a live researcher to wrap up, or launches one in synthesis mode when none is running. Use when results are good enough or the agent is stalling.
status
Show the current state of a research-agent run from .research/: liveness, cycles and budget, sub-problem and hypothesis statuses, eval trend, current focus, and branches. Read-only.
resume
Resume an interrupted research-agent run from .research/. Re-launches the researcher to reload its structured state and continue from the current focus, optionally on a named branch or with a new budget. Use after a session restart, crash, or explicit pause.