By jiahao-shao1
Automates research paper analysis, code review, project management, and daily documentation workflows by integrating Claude Code with arXiv, Notion, Obsidian, Git, and remote clusters.
Fetch BibTeX entries from arXiv / Semantic Scholar. Triggers: 'fetch bibtex', 'cite this paper', 'arXiv bibtex', '拉 bibtex', '获取引用', '引用这篇论文'. Not for citation format conversion (APA/MLA) or literature review.
cmux multi-pane / multi-agent terminal orchestration via cmux CLI (Bash cannot split panes or spawn agents). Use for parallel panes, splits, spawning Claude/Codex sub-agents, sending keys (incl. ctrl-c) between panes, reading pane output, sidebar updates, browser/markdown panes. Triggers: 'in parallel', 'split pane', 'spawn agent', 'fan out', 'browser pane', 'sidebar', '分屏', '并行', '开个 pane'. Not for plain bash or generic tmux.
Cross-model second opinion from OpenAI Codex on plans or code. Auto-detects plan vs code review, iterates up to 5 rounds. Use for architecture decisions, non-trivial refactors, critical config changes. Triggers: '/codex-review', 'codex review', 'second opinion', 'cross-check', '让 codex 看看', '交叉审查'. Not for trivial fixes or formatting.
Delegate a coding task (debug, implement feature, refactor) to OpenAI Codex CLI directly via codex exec. Faster than codex:rescue agent because it skips the Node companion runtime. Codex writes the code; Claude verifies. Use when you want Codex to actually do the work, not just review. Triggers: '/codex-task', 'codex task', 'let codex do', 'delegate to codex', 'codex implement', 'codex debug', '让 codex 干', '让 codex 实现', '让 codex 修', '丢给 codex', '外包给 codex'. Not for review or second opinion (use codex-review).
Audit project context management (CLAUDE.md, rules, knowledge) for progressive disclosure compliance. Checks that all knowledge files are reachable from rules, detects stale references, orphaned files, and CLAUDE.md index leakage. Triggers: 'audit context', 'check context', 'context hygiene', 'context audit', '检查 context', '审计上下文', 'knowledge 覆盖检查'. NOT for: code review, rule content correctness, or knowledge content quality.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
English | 中文
A collection of Claude Code skills for research workflow automation.
| Skill | Description |
|---|---|
| scholar-agent | Taste-aware research agent — personalized paper discovery via Scholar Inbox + hallucination-free deep reading via NotebookLM |
| cmux | Agent-friendly terminal built on Ghostty — multi-agent orchestration via split panes, spawn sub-Claude-Code instances, built-in browser, markdown preview, sidebar progress reporting |
| daily-summary | Daily work summary — aggregates Claude Code sessions, git commits, and Notion tasks into a timeline-style Chinese report |
| notion-lifeos | Notion life management — PARA method + Make Time journaling, with natural language task/note/journal CRUD via Notion API |
| web-fetcher | Web page → clean markdown with 5-layer fallback: Jina Reader → defuddle.md → markdown.new → OpenCLI (platform-specific with login state) → raw HTML |
| init-project | Initialize Claude Code project config — CLAUDE.md scaffolding, agent templates, and research profile setup |
| project-review | Project strategy panoramic review — auto-discover strategy docs and generate a 5-dimension snapshot (vision, roadmap, blockers, related work, next steps) |
| remote-cluster-agent | Remote GPU cluster operations — edit code locally, run commands remotely via Go daemon + rca CLI (exec / batch / cp / nodes), persistent SSH connection pool, node health monitoring |
| codex-review | Cross-model review — send your plan or code diff to OpenAI Codex for independent verification, iterative Claude↔Codex feedback until approved (max 5 rounds) |
| codex-task | Delegate a coding task (debug / implement / refactor) directly to OpenAI Codex CLI. Single-shot codex exec with gpt-5.5 + xhigh effort and full sandbox bypass by default; Claude verifies the diff after. Faster than codex:rescue (skips the Node companion runtime) |
| paper-analyzer | Deep critical paper analysis — causal chain methodology (现象→实验设置→归因→解法), NotebookLM-grounded reading, optional research framework mapping |
| paper-self-review | Paper paragraph self-review — three-axis check (logic jumps / repetition / premature method detail), v1 → v2 deterministic iteration |
| experiment-registry | ML experiment lifecycle management — structured YAML registry with CLI for registering experiments, recording benchmarks, comparing results, and tracking status |
| handoff | Session handoff summary — prints a structured context summary (status, decisions, pitfalls, next steps) directly in the conversation for seamless session continuity |
| sync-docs | Documentation sync checker — scans recent code changes and reports which docs (knowledge base, registry, CLAUDE.md, rules, README) need updating. Report only, no auto-modify |
| context-audit | Context management hygiene — audits the three-layer architecture (CLAUDE.md / rules / knowledge) for progressive disclosure compliance. Detects orphaned knowledge, stale references, and CLAUDE.md index leakage. Read-only |
| bibtex-fetch | Fetch correct BibTeX entries from arXiv (by ID) or Semantic Scholar (by title search). Batch fetch, custom bibkeys, exponential backoff. Zero dependencies (Python stdlib only) |
| todo-worker | Pick the next executable task from project-root TODO.md by priority + dependency, atomically claim it ([ ] → [~]), then execute. Explicit invocation only; canonical task-line format with four-state machine (todo / in-progress / done / blocked) |
| obsidian-brain | ⏸️ On Hold — Obsidian Second Brain with dual-zone vault. Paused in favor of enhancing notion-lifeos with reflection commands |
/plugin marketplace add jiahao-shao1/sjh-skills
/plugin install sjh-skills@sjh-skills
/reload-plugins
Plugin auto-updates from GitHub on startup. No manual sync needed.
Tell Codex:
Fetch and follow instructions from https://raw.githubusercontent.com/jiahao-shao1/sjh-skills/refs/heads/main/.codex/INSTALL.md
Or manually:
git clone https://github.com/jiahao-shao1/sjh-skills.git ~/.codex/sjh-skills
mkdir -p ~/.agents/skills
for skill in ~/.codex/sjh-skills/skills/*/; do
ln -sf "$skill" ~/.agents/skills/$(basename "$skill")
done
Detailed docs: .codex/INSTALL.md
npx skills add jiahao-shao1/sjh-skills --skill scholar-agent
npx skills add jiahao-shao1/sjh-skills --skill cmux
All skills at once:
npx claudepluginhub jiahao-shao1/sjh-skills --plugin sjh-skillsMake your AI agent code with your project's architecture, rules, and decisions.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Build and maintain an LLM-curated personal knowledge base in your project — Andrej Karpathy's LLM Wiki pattern, designed to scale to thousands of pages without becoming a context bottleneck. Now with an optional compiled graph layer for typed, provenance-backed relationships.
Connect to Atlassian products including Jira and Confluence. Search and create issues, access documentation, manage sprints, and integrate your development workflow with Atlassian's collaboration tools.
AI-powered wiki generator for code repositories. Generates comprehensive, Mermaid-rich documentation with dark-mode VitePress sites, onboarding guides, deep research, and source citations. Inspired by OpenDeepWiki and deepwiki-open.
20 modular skills for idiomatic Go — each under 225 lines, backed by 48 reference files, 8 automation scripts (all with --json, --limit, --force), and 4 asset templates. Covers error handling, naming, testing, concurrency, interfaces, generics, documentation, logging, performance, and more. Activates automatically with progressive disclosure and conditional cross-references.