AGENT SKILL

revise

Apply fixes from a multi-agent review report. Launches the reviser orchestrator, which triages the findings in .papermill/reviews/, dispatches section-writer specialists per finding domain, verifies each fix with the owning review specialist, and writes a revision report paired with the review. Closes the draft, review, revise loop. 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
revise
description
Apply fixes from a multi-agent review report. Launches the reviser orchestrator, which triages the findings in .papermill/reviews/, dispatches section-writer specialists per finding domain, verifies each fix with the owning review specialist, and writes a revision report paired with the review. Closes the draft, review, revise loop. Updates .papermill/state.md.
argument-hint
[path to review.md]

03 SKILL.md 正文

Revise From Review Findings

Launch the multi-agent revision system on an existing review report.

Step 1: Read Context

Read .papermill/state.md (Read tool) if it exists for the paper path, format, and the latest review_history entry. Work gracefully without it.

Step 2: Locate the Review Report

  • If a report path was supplied ($ARGUMENTS, or named in the request), use it.
  • Otherwise take the newest .papermill/reviews/*/review.md (Glob tool, sort by date directory).
  • If no report exists, stop and suggest /papermill:review first.

If the chosen review_history entry already has a revised date, tell the user and ask whether to re-revise against the same report or run a fresh review first.

Step 3: Confirm Scope

Ask the user (AskUserQuestion) which severities to address: all findings (default), Critical + Major only, or a custom selection. Mention how many findings the report contains per severity so the choice is informed.

Step 4: Launch the Reviser Orchestrator

Spawn the reviser agent (Agent tool, subagent_type: papermill:reviser) with:

<report>{path to review.md}</report>
<paper>{manuscript path from state.md or discovery}</paper>
<scope>{severity filter from Step 3}</scope>

Work through the findings with fix-then-verify. Write revision.md next
to the source review report.

Step 5: Present Results

Summarize from the agent's report: findings fixed and verified, decisions needed from the author, retry failures (reverted), and the build status. Point at the revision report path.

Step 5b: Resolve Decisions Needed

The reviser cannot ask questions itself (Claude Code removes AskUserQuestion from subagents), so it returns judgment calls as a decisions-needed YAML block. If that block is non-empty:

  1. Ask the author with AskUserQuestion, one question per decision, using each item's question, its options (label as the option, consequence as the description), and recommended listed first with "(Recommended)". Batch up to 4 questions per call; group decisions that touch the same section.

  2. Spawn the reviser again (Agent tool, subagent_type: papermill:reviser) with the same <report>, <paper>, and <scope> (so finding ids resolve against the same set), plus:

    <decisions>
    - id: F7
      choice: "Narrow the claim"
    </decisions>
    

    It fixes and verifies only those findings and appends them to the same revision.md.

  3. Fold its fixed-and-verified count into the Step 6 total. Decisions the author explicitly skips stay listed as open.

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:

Augment the review_history entry that matches the source report with:

    revised: "YYYY-MM-DD"
    findings_fixed: N        # fixed-and-verified count from the revision report

Append a timestamped note to the markdown body summarizing the revision pass.

Step 7: Suggest Next Steps

  • If everything in scope was fixed: suggest re-running /papermill:review to confirm the recommendation moves toward ready.
  • If the author skipped any decisions in Step 5b: list them as still open; re-running /papermill:revise on the same report picks them up.
  • If the build failed: fix the build before anything else.

04 同仓库更多 Skills

Abdulqudduz
Abdulqudduz/claude-anvil

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.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

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".

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

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.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

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.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

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.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

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.

★ 0