AI Skill Hub 推荐使用:智能代码阅读 是一款优质的Agent工作流。AI 综合评分 7.5 分,在同类工具中表现稳健。如果你正在寻找可靠的Agent工作流解决方案,这是一个值得深入了解的选择。
智能代码阅读 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
智能代码阅读 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:cargo install(推荐) cargo install tilth # 方式二:从源码编译 git clone https://github.com/jahala/tilth cd tilth cargo build --release # 二进制在 ./target/release/tilth
# 查看帮助 tilth --help # 基本运行 tilth [options] <input> # 详细使用说明请查阅文档 # https://github.com/jahala/tilth
# tilth 配置说明 # 查看配置选项 tilth --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export TILTH_CONFIG="/path/to/config.yml"
Smart code reading for humans and AI agents. Reduces cost per correct answer by 44% on Sonnet, 39% on Opus, and 38% on Haiku across 160 benchmark runs. (benchmarks)
tilth is what happens when you give ripgrep, tree-sitter, and cat a shared brain.
```bash $ tilth src/auth.ts
Rust. ~20,000 lines. No runtime dependencies.
grep-regex, grep-searcher) — fast content searchSearch runs definitions and usages in parallel via rayon::join. Callee resolution runs at expand time — extract callee names via tree-sitter queries, resolve against the source file's outline and imported files. Callers query uses the same tree-sitter patterns in reverse, walking the codebase with memchr SIMD pre-filtering for fast elimination.
The search output format is informed by wavelet multi-resolution (outline headers show line ranges for drill-down) and 1-hop callee expansion (expanded definitions resolve callees inline).
```bash cargo install tilth
→ [34] router.use('/api/protected/*', handleAuth); ```
Tree-sitter finds where symbols are defined — not just where strings appear. Each match shows its surrounding file structure so you know what you're looking at without a second read.
Expanded definitions include a callee footer (── calls ──) showing resolved callees with file, line range, and signature — the agent can follow call chains without separate searches for each callee.
tilth <path> # read file (outline if large)
tilth <path> --section 45-89 # exact line range
tilth <path> --section "## Foo" # markdown heading
tilth <path> --full # force full content
tilth <symbol> --scope <dir> # definitions + usages
tilth <symbol> --expand=5 # inline source for top 5 matches
tilth <symbol> --callers # find call sites (structural)
tilth <path> --deps # imports + dependents
tilth "TODO: fix" --scope <dir> # content search
tilth "/<regex>/" --scope <dir> # regex search
tilth "*.test.ts" --scope <dir> # glob files
tilth diff HEAD~1 # structural diff (function-level)
tilth --map --scope <dir> # codebase skeleton (CLI only)
--map is available in the CLI but not exposed as an MCP tool — benchmarks showed AI agents overused it, hurting accuracy.
智能代码阅读工具,提高效率
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
总体来看,智能代码阅读 是一款质量良好的Agent工作流,在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态,建议收藏备用,结合自身场景选择合适时机引入使用。
| 原始名称 | tilth |
| 原始描述 | 开源AI工作流:Smart(er) code reading for humans and AI agents. Reduces cost per correct answer。⭐245 · Rust |
| Topics | aiclirust |
| GitHub | https://github.com/jahala/tilth |
| License | MIT |
| 语言 | Rust |
收录时间:2026-05-29 · 更新时间:2026-05-30 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端