ml-paper-writing

name	ml-paper-writing
description	Use when writing or revising any ML/AI paper (conference or journal) to apply shared writing quality, citation verification, reviewer-aligned checks, and pre-submission gates.
allowed-tools	["read_file","write_file","str_replace","bash"]

[SKILL: ML PAPER COMMON WORKFLOW]

Activate when: user asks to write/revise/check any ML/AI paper, regardless of venue (conference or journal).

Purpose

This skill is the shared, venue-agnostic layer. Venue-specific skills should handle template/macros/page limits, while this skill handles writing quality, citation integrity, reviewer alignment, and submission checks.

Coverage Boundary

• This skill covers common writing/citation/review SOP for both conferences and journals.
• This skill does not provide journal-specific template setup (.cls/.sty/.bst, submission system steps, journal-only formatting policies).
• For journals, pair this skill with the journal's official author guide or a journal-specific venue skill.

Required Reference Files

Use these references from references/ as the source of truth:

1. writing-guide.md — paper narrative and section-level writing quality
2. citation-workflow.md — verified citation workflow and hallucination prevention
3. reviewer-guidelines.md — reviewer evaluation lenses and rebuttal-oriented checks
4. checklists.md — mandatory checklist gates before submission

Execution Workflow

Step 1: Clarify Scope

• Confirm paper type: conference or journal
• Confirm current stage: first draft / revision / camera-ready
• Confirm what is venue-specific vs common writing work

Step 2: Apply Writing Quality Workflow

• Use writing-guide.md to enforce:
  • clear contribution framing
  • precise claims and evidence alignment
  • section-level coherence (abstract, intro, method, results, limitations)

Step 3: Apply Citation Integrity Workflow

• Follow citation-workflow.md strictly
• Never fabricate citations from memory
• If citation cannot be verified, mark as explicit placeholder and report it

Step 4: Apply Reviewer Lens

• Use reviewer-guidelines.md to evaluate:
  • technical soundness
  • novelty/significance
  • clarity and reproducibility
• Add missing evidence/ablation/limitations where needed

Step 5: Run Submission Gates

• Use checklists.md and complete all applicable items
• If any required item is missing, block "ready to submit" status and list the gaps

Handoff Rule to Venue Skills

• Venue skills own:
  • template selection and setup (.cls/.sty/.bst)
  • anonymization mode and page limits
  • venue-specific required sections and package restrictions
• This skill owns common writing/citation/review/checklist quality gates.

Use both layers together: ml-paper-writing + target venue skill.