菜单
Reference for using Claude Code CLI to handle complex coding tasks. Use when Research-Claw needs to delegate multi-file projects, code review, refactoring, or agentic coding workflows to Anthropic's Claude Code. Trigger words: claude code, claude cli, delegate coding, code agent.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Main orchestration workflow for systematic literature research - search, evaluate, traverse, synthesize via the rp_* tools pipeline (Scopus + OpenAlex)
Collaboratively build and refine paper screening rubrics through brainstorming, test-driven development, and iterative feedback
Safely remove intermediate files from completed research sessions while preserving important data
Perform advanced search on CNKI with field filters like author, title, journal, date range, source category (SCI/EI/CSSCI/北大核心). Use when user needs precise filtered search beyond simple keywords.
Download a paper PDF/CAJ from CNKI. Requires user to be logged in. Use when user wants to download a specific paper.
Export paper from CNKI and push to Zotero, or save as RIS file. Use when user wants to save a paper to Zotero or export citation data.
| name | claude-code |
| description | Reference for using Claude Code CLI to handle complex coding tasks. Use when Research-Claw needs to delegate multi-file projects, code review, refactoring, or agentic coding workflows to Anthropic's Claude Code. Trigger words: claude code, claude cli, delegate coding, code agent. |
Agentic coding tool by Anthropic. Reads, edits, and runs code with full project context. Supports non-interactive print mode for programmatic use.
which claude # should return a path
claude --version
npm install -g @anthropic-ai/claude-code
Requires authentication: run claude auth login (Anthropic account or --console for API billing).
For programmatic invocation from Research-Claw, use -p (print mode):
# Basic non-interactive execution
claude -p "Refactor the auth module to use async/await"
# Pipe content for analysis
cat src/utils.ts | claude -p "Find bugs in this file"
# JSON output for parsing
claude -p --output-format json "List all TODO comments in the project"
# Skip permission prompts (for fully automated pipelines)
claude -p --dangerously-skip-permissions "Fix all lint errors"
# Budget and turn limits
claude -p --max-turns 5 --max-budget-usd 2.00 "Add unit tests for the auth module"
# Model selection
claude -p --model sonnet "Explain this codebase architecture"
| Flag | Description |
|---|---|
-p, --print | Non-interactive mode: run task, print result, exit |
--model <model> | Set model (sonnet, opus, or full model ID) |
--output-format <fmt> | text / json / stream-json |
--max-turns <n> | Limit agentic turns (print mode) |
--max-budget-usd <n> | Spending cap in USD (print mode) |
--dangerously-skip-permissions | Skip all permission prompts |
--allowedTools <tools> | Auto-approve specific tools, e.g. "Bash(git *)" "Read" |
--tools <tools> | Restrict available tools, e.g. "Bash,Read,Edit" |
--add-dir <path> | Grant access to additional directories |
--append-system-prompt <text> | Append to default system prompt |
-c, --continue | Continue most recent conversation |
-r, --resume <id> | Resume session by ID or name |
--effort <level> | low / medium / high / max |
# Delegate a complex coding task
exec("claude -p --dangerously-skip-permissions --max-turns 10 'Implement the feature described in TODO.md'")
# Read-only analysis with budget cap
exec("claude -p --tools 'Read' --max-budget-usd 1.00 'Review code quality and list issues'")
# JSON output for structured results
exec("claude -p --output-format json 'List all exported functions in src/'")