papermill/skills/draft/SKILL.md
Launch a multi-agent writing system with specialist writers for literature, formal content, methodology, and results, orchestrated by a lead author. Produces a manuscript draft and section-level artifacts in .papermill/drafts/. Updates .papermill/state.md.
npx skillsauth add queelius/claude-anvil draftInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Launch a multi-agent system to draft a research paper. The writing system uses specialist writers for different content types (related work, mathematical content, methodology, results), coordinated by a lead author who plans the paper, integrates sections, and writes the bookend sections (abstract, introduction, conclusion).
Read .papermill/state.md (Read tool) for:
If .papermill/state.md does not exist, warn the user:
I can draft without a state file, but results will be much better if you first run:
/papermill:init— to set up the project/papermill:thesis— to crystallize the central claim/papermill:outline— to design the paper structureWant to proceed anyway, or set up the project first?
If no outline exists but a thesis does, the writer orchestrator will create an outline as its first step — but an author-approved outline produces better results.
Scan the project for existing materials:
*.tex, *.md, *.Rmd) — is there already a partial draft?*.bib) — existing referencescode/, scripts/, data/, results/) — evidence to incorporateimages/, figures/, fig/, *.png, *.pdf in paper directories)Summarize what exists to the user:
Project inventory:
- Thesis: [yes/no — quote if yes]
- Outline: [yes/no — summarize sections if yes]
- Existing draft: [yes/no — describe what exists]
- Bibliography: [N entries in file.bib]
- Code/data: [brief description]
- Figures: [N figures available]
Ask the user what they want drafted:
I can draft:
- Full paper — all sections from abstract to conclusion
- Specific sections — choose which sections to write
- Extension — add new sections/content to an existing draft
Which approach? And is there anything specific I should know about the paper's direction?
If the user chooses specific sections, note which ones. If they choose extension, identify what already exists and what needs adding.
Launch the writer agent (Task tool with subagent_type: "papermill:writer").
Pass the agent:
.papermill/state.md (if it exists)The agent will:
After the agent completes, read the manuscript file (Read tool).
Present a summary to the user:
Draft Complete
Manuscript: [path to main file] Sections drafted: [list of sections with specialist who wrote each] Word/page count: [approximate] Build status: [compiles cleanly / has N warnings / needs fixes]
Section drafts and writing plan:
.papermill/drafts/YYYY-MM-DD/The draft is a starting point, not a final product. Key areas to review:
- [Most important thing to check — e.g., "Proof in Section 4 may need verification"]
- [Second priority — e.g., "Results section has placeholders for experimental data"]
- [Third priority — e.g., "Related work positioning could be sharpened"]
Then ask: "Would you like to review specific sections, or run /papermill:review for a systematic evaluation?"
Update .papermill/state.md (Edit tool):
stage to drafting or draft-complete- YYYY-MM-DD (draft): Multi-agent draft complete. Sections: [list]. Specialists used: [list]. Placeholders: [count if any].
Based on the draft's state, suggest the most relevant next step:
/papermill:review for systematic multi-agent evaluation/papermill:draft targeting specific sections, or edit directly."/papermill:polish for pre-submission preparation/papermill:proof to check mathematical contentdevelopment
Force a research-agent run to conclude. Launches the researcher in synthesis mode: it reads state.md and log.md, writes .research/synthesis.md with outcome, key findings, failed approaches, open questions, and recommendations, then exits. Use when current results are good enough or the agent is stalling.
data-ai
Show the current state of an in-flight research-agent run from .research/state.md, log.md, and attempts/. Read-only summary of cycles, sub-problems, hypothesis statuses, eval trend, and current focus.
testing
Resume an interrupted research-agent run. Re-launches the researcher with instructions to read .research/state.md and log.md, reorient, and continue from the documented current focus. Use after a context compression, session restart, or explicit pause.
tools
When and how to use the repoindex plugin surface (MCP tools, agents, slash commands) for collection queries, release prep, activity summaries, and tag discipline. Use when users ask repoindex questions, mention their repo catalog, or want to know which repoindex tool fits their task.