Assist with Colibri: pure-C LLM inference engine for running GLM-5.2 (744B MoE) on consumer machines with ~25 GB RAM. Use when setting up, building, converting models, running inference, configuring expert streaming and caching, optimizing speculative decoding (MTP), GPU integration, and integrating Colibri into production pipelines. Includes build setup, model download & conversion, chat/inference modes, performance tuning, and API integration patterns.

2026-07-1237

prompts-chat

akillness/jeo-skills

Discover and apply curated prompts from the prompts.chat collection to optimize AI interactions. Use when refining prompt engineering, finding domain-specific prompt templates, improving response quality, or building prompt-based workflows. Triggers on: prompt optimization, prompt templates, prompt engineering, prompt library, curated prompts, prompt discovery, and AI prompt patterns.

2026-07-1237

vox-director

akillness/jeo-skills

Turn ONE topic into a finished Vox-style paper-collage explainer / ad video, end to end on the Atlas Cloud API + local ffmpeg — script, collage keyframes, motion, voice-over, music, captions, all automated. Use this whenever the user wants a "Vox style" video, a paper/torn-paper collage animation, a "motion collage", a narrated explainer or short ad built from AI-generated collage posters, a scrapbook-style tribute, or wants to turn a topic / product / person into a punchy narrated collage video — even if they don't say the word "Vox". Also use when reproducing Stav Zilber / rom1trs / Higgsfield-style collage ad workflows, or when the user asks for a motion collage or a scrapbook-style tribute. Triggers: "vox video", "collage video", "motion collage", "paper collage explainer", "make a collage ad", "turn this topic into a collage video".

2026-07-1237

motion-previs-studio

akillness/jeo-skills

Assist with Motion Previs Studio v4: a cross-platform desktop app for AI-film previsualization. Use when setting up, configuring, troubleshooting, or extending motion-previs-studio for pose extraction, depth mapping, camera motion solving, control layer export, and bundle production for AI-video workflows (Seedance, ComfyUI, Blender, Runway, Kling). Includes build setup, feature integration, UI/logic debugging, and export pipeline optimization.

2026-07-1237

lapian-notes

akillness/jeo-skills

Work with Lapian Notes / 拉片笔记 (github.com/bkingfilm/lapian-notes) — a local- first React/Vite tool that turns a film into an editable shot-by-shot study notebook: local frame extraction, AI-assisted structure analysis (bring your own AI, no API key required), story-line swimlane timeline, structure tree, and audience-emotion curve. Use when the user asks about Lapian Notes, "拉片笔记", "拉片" (shot-by-shot film analysis) tooling, cloning/running this repo (npm run dev, run.bat/run.command), the AI-analysis-package (ZIP) round-trip workflow, or contributing a PR to lapian-notes. Not for generic video editing (use `opencut` for that) or generic film-analysis theory unrelated to this codebase.

2026-07-1237

tokhub

akillness/jeo-skills

Set up, run, and contribute to TokHub (github.com/yaojingang/TokHub) — an open-source AI API relay monitoring, recommendation, and OpenAI-compatible gateway system with L1/L2/L3 channel health probing, usage metering, alerts, audit, and Docker self-hosting. Use when the user asks about TokHub, "AI API 中转站监控", cloning/running the Go + React monorepo (TOKHUB_ROLE, sqlc, TimescaleDB, NATS), the L1/L2/L3 probe algorithm, the OpenAI-compatible `/gateway/v1/*` endpoint, or contributing a PR to TokHub. Do not use for connecting a running agent to a live TokHub instance's own API (that is covered by the project's own bundled `agent-skills/tokhub` skill inside the TokHub repo, not this one).

2026-07-1237

name	academic-research
description	>
compatibility	>
allowed-tools	Bash Read Write Edit Glob Grep WebFetch
metadata	{"tags":"academic-research, deep-research, paper-writing, peer-review, literature-review, systematic-review, citation-check, research-pipeline, scholarly-publishing, ars","version":"1.0.0","source":"https://github.com/Imbad0202/academic-research-skills","upstream_version":"3.13.0","license":"CC-BY-NC-4.0"}

academic-research

A routing-first front door for the full Academic Research Skills (ARS) suite — 4 pipelines, 27 modes, spanning the complete research-to-publication lifecycle.

AI is copilot, not pilot. ARS handles the grunt work (hunting references, verifying citations, checking logical consistency, formatting). You handle the parts that require your brain: defining the question, choosing the method, interpreting results, and writing the sentence after "I argue that."

Read these pipeline references before executing:

Plugin Installation

# Full upstream suite (all 4 skills, 27 modes — recommended for heavy use)
claude plugin marketplace add Imbad0202/academic-research-skills

# Or install this routing skill only via jeo-skills
npx skills add https://github.com/akillness/jeo-skills --skill academic-research

When to use this skill

Conducting rigorous literature research on a topic (→ deep-research pipeline)
Writing a new academic paper from scratch or guided outline (→ academic-paper pipeline)
Reviewing or evaluating an existing paper (→ academic-paper-reviewer pipeline)
Orchestrating the full research-to-publication workflow end-to-end (→ academic-pipeline)
Fact-checking claims in a paper or source (→ deep-research / fact-check mode)
Preparing for peer review, rebuttal, or journal revision (→ academic-paper / revision modes)
Running a PRISMA 2020 systematic review (→ deep-research / systematic-review mode)
Generating a bilingual abstract, converting citations, checking AI disclosure requirements

When not to use this skill

The request is a code documentation or API reference → use technical-writing
The request is an ML experiment loop (training/eval) → use autoresearch
The request is a general writing task (blogs, marketing, newsletters) → use marketing-automation
The request needs autonomous agent ML training → use autoresearch
The user already has a paper draft and just needs copy editing → use technical-writing or user-guide-writing

Required intake packet

Before routing, identify:

Pipeline — which of the 4 pipelines fits (deep-research / academic-paper / academic-paper-reviewer / academic-pipeline)
Mode — which specific mode within that pipeline (see routing table below)
Topic or input — the research topic, paper text, or reviewer comments to work from
Oversight preference — Very High (guided/Socratic) → High (key decisions) → Medium (structured) → Low (template)
Format / venue — APA 7.0, IEEE, LaTeX, NeurIPS, ICLR, CVPR, or journal target

Pipeline & Mode Routing Table

What the user says	Pipeline	Mode	Oversight
"research [topic]", "deep research", "academic analysis"	deep-research	full	High
"quick brief", "30 minute summary"	deep-research	quick	Medium
"review this paper's research quality"	deep-research	review	High
"literature review", "annotated bibliography"	deep-research	lit-review	Medium
"WHY HOW WHAT papers", "3W scan", "compare papers"	deep-research	three-way-scan	Low
"verify claims", "fact-check", "evidence verification"	deep-research	fact-check	Medium
"guide my research", "help me think through"	deep-research	socratic	Very High
"systematic review", "meta-analysis", "PRISMA"	deep-research	systematic-review	Medium
"write a paper on X", "academic paper", "research paper"	academic-paper	full	High
"guide me through writing", "help me plan my paper"	academic-paper	plan	Very High
"build a paper outline"	academic-paper	outline-only	High
"I have reviewer comments", "revise my paper"	academic-paper	revision	High
"parse reviewer comments into a roadmap"	academic-paper	revision-coach	Medium
"write an abstract"	academic-paper	abstract-only	Medium
"turn this into a literature review paper"	academic-paper	lit-review	Medium
"convert to LaTeX", "convert citations to IEEE"	academic-paper	format-convert	Medium
"check citations"	academic-paper	citation-check	Medium
"generate AI disclosure statement for NeurIPS"	academic-paper	disclosure	Medium
"audit my rebuttal draft against reviews"	academic-paper	rebuttal-audit	Medium
"review this paper" → full EIC + R1/R2/R3 + Devil's Advocate	academic-paper-reviewer	full	High
"quick assessment of this paper"	academic-paper-reviewer	quick	Low
"guide me to improve this paper"	academic-paper-reviewer	guided	Very High
"check the methodology"	academic-paper-reviewer	methodology-focus	Medium
"verify the revisions"	academic-paper-reviewer	re-review	Medium
"calibrate this reviewer against my gold set"	academic-paper-reviewer	calibration	Medium
"I want to write a complete research paper"	academic-pipeline	full (10-stage)	Very High
"continue pipeline from reset boundary"	academic-pipeline	resume_from_passport	High

Instructions

Step 1: Pick the pipeline

deep-research        → topic investigation, literature synthesis, fact-checking, PRISMA
academic-paper       → write, outline, revise, abstract, format, rebuttal, disclosure
academic-paper-reviewer → evaluate, review, calibrate reviewer quality
academic-pipeline    → full orchestrated 10-stage research → paper → review → revise → finalize

If ambiguous between deep-research and academic-paper: the user doing discovery → deep-research; the user writing or revising a document → academic-paper.

Step 2: Confirm the mode and input

State the chosen pipeline and mode explicitly before proceeding
For socratic or plan modes: begin Socratic dialogue immediately (one question at a time)
For full or revision modes: confirm the topic/paper and output format first
For citation-check or fact-check: confirm which claims or sections to target

Step 3: Execute with human-in-the-loop

Follow the pipeline reference for the chosen mode:

deep-research: Read references/deep-research-pipeline.md
academic-paper: Read references/academic-paper-pipeline.md
academic-paper-reviewer: Read references/academic-paper-reviewer-pipeline.md
academic-pipeline (10-stage): Read references/academic-pipeline-stages.md

Checkpoints marked [USER CHECKPOINT] require explicit user confirmation before continuing. Never skip them.

Step 4: Return a structured output packet

Every completed mode should return:

Pipeline:     <which pipeline>
Mode:         <which mode>
Topic/Input:  <research topic or paper title>
Output:       <the primary artifact — report / draft / review / outline>
Citations:    <APA 7.0 / IEEE / inline — as applicable>
Integrity:    <claim verification status if fact-check or citation-check was run>
Next step:    <recommended follow-up mode or pipeline>

Data access and integrity principles

data_access_level: verified_only — ARS never uses unverified claims in outputs
task_type: open-ended — all modes are open-ended; ARS does not run experiments
Citation hallucination guard: three-index triangulation (Semantic Scholar + OpenAlex + Crossref) in academic-pipeline Stage 6/7
Experiment provenance: claims backed by experiments must be declared in the Material Passport; ARS audits alignment (ALIGNED / OVERSTATED / NOT_SUPPORTED_BY_PROVENANCE)
Style Calibration: academic-paper plan mode can learn your voice from past work samples

Route-out map

If the user needs…	Route to
ML training experiment loop	`autoresearch`
Prompt / skill eval pipeline	`skill-autoresearch`
Engineering design doc or ADR	`technical-writing`
API portal or SDK reference	`api-documentation`
Karpathy-style ML research	`autoresearch`
General content / blog / newsletter	`marketing-automation`
Code documentation	`technical-writing`