Academic Research Skills for Claude Code: A Human-in-the-Loop Pipeline for Paper Writing
Academic Research Skills (ARS) for Claude Code is a plugin that supports the full research-to-publication pipeline: research → write → review → revise → finalize. It is designed as a human-in-the-loop system, explicitly rejecting full automation. The tool handles grunt work — reference hunting, citation formatting, data verification, logical consistency checks — while the researcher retains control over question definition, method selection, interpretation, and the core argument.
Quick Install (Claude Code v3.7.0+)
/plugin marketplace add Imbad0202/academic-research-skills
/plugin install academic-research-skillsVerify with /ars-plan to start a Socratic dialogue mapping out chapter structure, or /ars-lit-review "your topic" for a single-shot literature review.
Why Human-in-the-Loop?
ARS cites Lu et al. (2026, Nature 651:914-919) who built The AI Scientist — the first fully autonomous AI research system to publish a paper through blind peer review at a top ML venue (ICLR 2025 workshop, score 6.33/10 vs workshop average 4.87). Their Limitations section lists failure modes: implementation bugs, hallucinated results, shortcut reliance, bug-as-insight reframing, methodology fabrication, frame-lock, and citation hallucinations. ARS is built on the premise that a human researcher augmented by AI avoids these failure modes better than either alone.
Integrity Gates & Calibration
Stage 2.5 and Stage 4.5 run a 7-mode blocking checklist (see academic-pipeline/references/ai_research_failure_modes.md). The reviewer offers an opt-in calibration mode that measures its own false negative rate / false positive rate against a user-supplied gold set.
Features
- Style Calibration learns your voice from past work.
- Writing Quality Check catches patterns that make prose feel machine-generated.
- Semantic Scholar API verification (inspired by PaperOrchestra, Song, Song, Pfister & Yoon, 2026, Google).
- Anti-leakage protocol, VLM figure verification, and score trajectory tracking.
Architecture
Full pipeline documentation is in docs/ARCHITECTURE.md, including flow diagram, stage-by-stage matrix, data-access flow, skill dependency graph, quality gates, and mode list. For DOCX output, pandoc is required; for APA 7.0 PDF, tectonic + Source Han Serif TC font (Markdown output works without either).
📖 Read the full source: HN AI Agents
👀 See Also
Rift: A Better Alternative to Git Worktrees with Instant Copy-on-Write Snapshots
Rift uses btrfs or APFS snapshots to create instant, space-efficient copies of Git repositories. Initialization, creation, and listing via CLI or JavaScript FFI.
Testreel: Programmatic Demo Video Generation with Claude Code
Testreel is an npm package that generates polished product demo videos from JSON, YAML, or Playwright interaction descriptions. It creates webm/mp4/gif videos with cursor overlays, click ripples, and gradient backgrounds.
Unofficial Ultrahuman Ring MCP Server for AI Agent Integration
A community-developed MCP server wraps the Ultrahuman Partner API, allowing AI coding agents to directly access ring and CGM metrics like sleep, HRV, glucose, and recovery scores via structured data calls.
Echo-TTS Ported to Apple Silicon with MLX for Native TTS with Voice Cloning
Echo-TTS, a 2.4B parameter diffusion text-to-speech model with voice cloning, has been ported from CUDA to run natively on Apple M-series silicon using MLX. On a base 16GB M4 Mac mini, a 5-second voice clone takes about 10 seconds to generate, while 30-second clones take about 60 seconds.