agentsclimarketplace

Academic paper writer

Skill joshua-zyy/academic-paper-writer/skills/academic-paper-writer

面向 CS / AI / ML 领域的证据驱动、分节推进的论文写作 Agent Skill。

Install
npx -y skills add joshua-zyy/academic-paper-writer --skill academic-paper-writer

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Use when writing or revising CS/AI/ML papers from research notes, code repositories, local literature libraries, target venue requirements, or section-level drafting requests. Triggers include 写论文, 初稿, paper draft, write introduction, write method, full paper outline, section-by-section drafting, 证据闭环, 论文起草, 从零写论文, 逐节写作.

SKILL.md

4.2 KB, as published. Nobody here has run it

Academic Paper Writer (Core Orchestrator)

证据闭环型、分节推进的论文编排代理。协调六个子技能:Step 0→1→2→3→4→5→6→7→8→9→10→11→12。

Router Protocol

  1. Read manifest.yaml. It declares always_load files, axes, and references.on_demand.
  2. Read every file listed under always_load. These are the skill's binding rules — not reference material.
  3. Apply the loaded material as constraints:
    • stance.md defines core rules, sub-skill boundary, and scope.
    • red-lines.md defines absolute prohibitions. Do not negotiate these.
    • output-contract.md defines file outputs, deliverables, and completion criteria.
    • anti-patterns.md defines known failure modes and their correct alternatives.
    • workflow.md defines the Step 0-12 sequence summary.
    • ../shared/core/ policies define evidence, non-invention, and output boundaries.
  4. Detect mode, paper_type, section, and language using manifest axes.
  5. Echo the detected mode to the user before executing.
  6. Reach for references/ only when the manifest's references.on_demand condition is satisfied.

Modes

ModePurpose
full-paper-planningFrom-scratch full paper (balanced spectrum)
section-draftingSingle section, narrowed evidence scope (balanced)
section-revisionLocal evidence audit + rewrite (fidelity spectrum)
related-work-or-citation-passDelegate to academic-citation (fidelity spectrum)
experiment-evidence-passDelegate to academic-experiments (fidelity spectrum)

Sub-Skill Dispatch

TaskSub-SkillStep
Venue researchacademic-venue-researchStep 4
Evidence auditProbe agentStep 5
Citation searchacademic-citationStep 6
Experiment auditacademic-experimentsStep 7
Figure generationacademic-figureStep 9.4
Prose polishingacademic-polishingStep 9.6 (internal)
Draft reviewacademic-reviserStep 9.5 / 9.8

Push Modes

ModeBehavior
auto (default)Auto-advance after Verification; brief progress summary only
step-by-stepPause after each section for user confirmation

Switch anytime via user request.

Hard Gates (A-E)

GateTriggerConditionFailure
E: VenueStep 1→5若用户提供本地 PDF 文献库,Step 2 已验证 MD 目录;venue-brief.md existsBlocked
A: EvidenceStep 5→9>= 1 usable evidenceDegradation or blocked
B: Section Citation ReadinessStep 6→9Current section has VERIFIED citations, Citation-to-Claim Map, or explicit [REF_NEEDED] debt for key claimsSection-dependent: Intro/RW blocked, Method/Results may proceed with tracked debt
C: VerificationStep 9.8→10All hard debts closed + thin_draft=nopassed/failed/blocked
D: Full-Paper Citation CountStep 11→outputTotal >= min_citationsWarn, allow retry

9.1 Pre-Draft Probe Rules

SectionProbes to DispatchStrategy
Introductionexisting_material + local lit deep search + external lit searchMust parallel (3 Tasks)
Related Workexisting_material + local lit deep search + external lit searchMust parallel (3 Tasks)
Methodcode_structure (Module Cards + tensor shapes + forward) + preprocessingMust parallel
Experimental Setupexperiment_setup (hyperparams, dataset split, demographics)Single probe
Results / Ablationexperiment_results (main results, baselines, ablation)Single probe
Discussioninterpretability (interpretability results, network analysis)Single probe

Default Section Queue

Abstract is post-posed, not in initial queue. Default: Introduction → Related Work → Method → Experimental Setup → Results → Discussion → Conclusion → Abstract.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.