[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-83120":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":17,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":18,"rankGlobal":10,"rankLanguage":10,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":10,"pushedAt":38,"updatedAt":39,"readmeContent":40,"aiSummary":41,"trendingCount":16,"starSnapshotCount":16,"syncStatus":42,"lastSyncTime":43,"discoverSource":44},83120,"Flowtrace","AIScientists-Dev\u002FFlowtrace","AIScientists-Dev","Run a task with AI as a flow of steps you keep, reuse, and refine, not a one-off chat.","https:\u002F\u002Fmorphmind.ai",null,"TypeScript",129,10,57,3,0,32,62.32,"MIT License",false,"main",true,[24,25,26,27,28,29,30,31,32,33,34,35,36,37],"agent-skills","agentic-ai","ai","ai-agent","ai-agents","anthropic","claude","claude-code","claude-skills","cli","codex","cursor","developer-tools","llm","2026-06-09 14:47:07","2026-06-10 19:51:58","\u003Cdiv align=\"center\">\n\n\u003Cimg src=\"docs\u002Fassets\u002Fhero.png\" width=\"440\" alt=\"Flowtrace\">\n\n# Flowtrace\n\n**Turn your agent's work into a transparent, reusable, evolving trace.**\n\nStructure the agent reasons through, and you can step in anywhere.\n\n[![MIT License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-MIT-blue?style=flat-square)](.\u002FLICENSE) [![Homepage](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FHomepage-morphmind.ai-lightgrey?style=flat-square)](https:\u002F\u002Fmorphmind.ai) [![Discord](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDiscord-Join-5865F2?style=flat-square&logo=discord&logoColor=white)](https:\u002F\u002Fdiscord.gg\u002Fx9mtbMEx) [![X](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FX-Follow-000000?style=flat-square&logo=x&logoColor=white)](https:\u002F\u002Fx.com\u002Fmorphmind__ai?s=11)\n\n[**What it does**](#what-it-does) · [**Get started**](#get-started) · [**Examples**](#examples) · [**Docs**](docs\u002Ftrace\u002FREADME.md)\n\n**English** · [简体中文](docs\u002FREADME.zh-CN.md)\n\n\u003C\u002Fdiv>\n\n---\n\nFlowtrace turns a task you run with your agent into a trace: a flow of steps the agent works through one at a time, instead of a single stream of text. Each step leaves its output on disk to open and check. It works with the agent and skills you already use, including Claude Code, Codex, and Cursor.\n\nOne real example: a buy-or-sell decision that ends in a fixed-format, citable PDF.\n\n\u003Cdiv align=\"center\">\n\u003Ctable>\u003Ctr>\n\u003Ctd align=\"center\" valign=\"top\">\u003Cimg src=\"docs\u002Fassets\u002Fexamples\u002Fnvda-decision.png\" height=\"340\" alt=\"A trace shown as a flow: ingest lanes fan into a synthesized thesis, then position sizing and risk controls, then a research-note report\">\u003Cbr>\u003Csub>The flow\u003C\u002Fsub>\u003C\u002Ftd>\n\u003Ctd align=\"center\" valign=\"top\">\u003Cimg src=\"docs\u002Fassets\u002Fexamples\u002Fnvda-decision-pdf.png\" height=\"340\" alt=\"The deliverable, a fixed-format research-note PDF\">\u003Cbr>\u003Csub>The deliverable\u003C\u002Fsub>\u003C\u002Ftd>\n\u003C\u002Ftr>\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\u003Ca href=\"docs\u002Fassets\u002Fexamples\u002Fnvda-decision.pdf\">\u003Cstrong>Read the full research-note PDF\u003C\u002Fstrong>\u003C\u002Fa>\u003C\u002Fp>\n\n## What it does\n\n**Transparent.** The whole task is a visible flow of steps, and each one shows what it did.\n\n**Grounded.** Every result can be checked, because the agent works one step at a time and each step leaves real files on disk. A reported result points back to those files, not to a claim.\n\n**Steerable.** The work is a structure to point at, not a conversation to scroll. A change applies to a single step, and whatever depends on it follows.\n\n**Traceable.** Nothing is ever lost. Every change is a git commit, so each run keeps a full audit trail: a new approach never overwrites the one that worked, and any earlier version can be brought back.\n\n**Reusable.** Once a trace exists, the next task of the same kind runs the same steps on new inputs. The method is reused, not rebuilt.\n\n**Evolving.** A step improved once carries into every later run. The trace gets better the more it runs.\n\n## Get started\n\nThe fast path is to hand the repo to an agent. Point a coding agent (Claude Code, Codex, Cursor) at this folder and say:\n\n> _\"Install Flowtrace and run the tailored-resume example.\"_\n\nIt installs the CLI, builds a real trace at `~\u002Ftraces\u002Ftailored-resume\u002F`, and opens the web view at `http:\u002F\u002Flocalhost:3000`, where the flow lights up step by step.\n\nTwo ways to get a trace:\n\n- **Try a reference.** Each example ships as a builder that creates a real trace folder and walks one full run.\n\n  ```bash\n  bash scripts\u002Fexamples\u002Ftailored-resume\u002Fbuild.sh   # → ~\u002Ftraces\u002Ftailored-resume\u002F\n  flowtrace serve                                  # → http:\u002F\u002Flocalhost:3000\n  ```\n\n- **Make your own.** The `make-trace` skill turns any source (a `SKILL.md`, a runbook, a chat log, a finished task) into a trace. Copy `skills\u002Fmake-trace\u002F` into the agent's skills directory and run `\u002Fmake-trace`.\n\nA run is steerable: stop at any step, change it, and the steps that depend on it re-run while the rest stay put.\n\n\u003Cdetails>\n\u003Csummary>Install by hand\u003C\u002Fsummary>\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FAIScientists-Dev\u002Fflowtrace.git\ncd flowtrace\n.\u002Fscripts\u002Finstall.sh        # builds + symlinks flowtrace to ~\u002F.local\u002Fbin\u002F\n```\n\nUpdate with `git pull && .\u002Fscripts\u002Finstall.sh`. Override the symlink target with `INSTALL_DIR=…`. Building from source or contributing? See [CONTRIBUTING.md](.\u002FCONTRIBUTING.md).\n\n\u003C\u002Fdetails>\n\n## Examples\n\n**Nine examples** built from popular open-source skills, spanning different domains — open any one for its flow and a one-command demo in the [examples gallery](docs\u002FEXAMPLES.md):\n\n\u003Cdiv align=\"center\">\n\u003Ctable>\u003Ctr>\n\u003Ctd align=\"center\" valign=\"top\">\u003Ca href=\"docs\u002FEXAMPLES.md#saas-dd\">\u003Cimg src=\"docs\u002Fassets\u002Fexamples\u002Ffeat-saas-dd.png\" height=\"240\" alt=\"SaaS acquisition due diligence flow\">\u003C\u002Fa>\u003Cbr>\u003Csub>\u003Ca href=\"docs\u002FEXAMPLES.md#saas-dd\">SaaS due diligence\u003C\u002Fa>\u003C\u002Fsub>\u003C\u002Ftd>\n\u003Ctd align=\"center\" valign=\"top\">\u003Ca href=\"docs\u002FEXAMPLES.md#security-cicd\">\u003Cimg src=\"docs\u002Fassets\u002Fexamples\u002Ffeat-security-cicd.png\" height=\"240\" alt=\"security CI\u002FCD pipeline flow\">\u003C\u002Fa>\u003Cbr>\u003Csub>\u003Ca href=\"docs\u002FEXAMPLES.md#security-cicd\">Security CI\u002FCD gate\u003C\u002Fa>\u003C\u002Fsub>\u003C\u002Ftd>\n\u003Ctd align=\"center\" valign=\"top\">\u003Ca href=\"docs\u002FEXAMPLES.md#distill-mind\">\u003Cimg src=\"docs\u002Fassets\u002Fexamples\u002Ffeat-distill-mind.png\" height=\"240\" alt=\"distill a mind into a skill flow\">\u003C\u002Fa>\u003Cbr>\u003Csub>\u003Ca href=\"docs\u002FEXAMPLES.md#distill-mind\">Distill a mind into a skill\u003C\u002Fa>\u003C\u002Fsub>\u003C\u002Ftd>\n\u003C\u002Ftr>\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\nPlus six more:\n\n- 📄 Career — [Tailored Résumé Generator](docs\u002FEXAMPLES.md#tailored-resume)\n- 💰 Investing — [Comprehensive Stock Analysis](docs\u002FEXAMPLES.md#nvda-decision)\n- ✍️ Research \u002F writing — [Industry Deep-Dive Report](docs\u002FEXAMPLES.md#research-writer)\n- 🐛 Software engineering — [Bug-Fix Learning Loop](docs\u002FEXAMPLES.md#swe-bugfix)\n- 📈 Growth \u002F marketing — [Weekly Paid-Ads Optimization](docs\u002FEXAMPLES.md#paid-ads)\n- 🖼 Design \u002F decks — [Talk → Magazine Slide Deck](docs\u002FEXAMPLES.md#talk-to-deck)\n\n## Documentation\n\nA trace is one git repository. `trace.json` declares the steps, their dependencies, and the final deliverable. Each run lives under `runs\u002F\u003Crun_id>\u002F`:\n\n```\n\u003Ctrace_root>\u002F\n├─ .git\u002F                                    standard git repo, the audit trail\n├─ trace.json                              the static plan (steps + deliverable)\n├─ scripts\u002F                                 shared code used by 2+ steps\n├─ resources\u002F                               shared static material (refs, papers, master data)\n├─ steps\u002F\u003Cstep_id>\u002F\n│  ├─ STEP.md                               per-step contract + impl hints\n│  ├─ scripts\u002F                              step-local code\n│  └─ resources\u002F                            step-local material (figures, PDFs, fixtures)\n└─ runs\u002F\u003Crun_id>\u002F\n   ├─ state.json                            run status (sole source of truth)\n   ├─ replies\u002FNNNN.json                     append-only structured-output stream\n   └─ \u003Cstep_id>\u002F                            run-time files (assets + scratch)\n```\n\nThe same two-name convention (`scripts\u002F` for code that runs, `resources\u002F` for static material that doesn't) appears at both the trace root and inside each step. Anything reused across 2+ steps belongs at the trace root; single-step material stays inside the step folder. `STEP.md` references either with relative paths.\n\nEvery CLI write makes one git commit, scoped to exactly the paths it declares: `state.json` plus any `--asset` paths, or the new reply file plus its cited evidence paths. Scratch files stay untracked. The git history is the audit trail, and the UI can time-travel through it.\n\nSteps pass data through files, not parameters: each step writes its output, and a downstream step reads it.\n\n| To learn | Read |\n|---|---|\n| The idea, in depth | [PHILOSOPHY.md](docs\u002Ftrace\u002FPHILOSOPHY.md) |\n| Driving a trace as an agent | [docs\u002Ftrace\u002FCLI.md](docs\u002Ftrace\u002FCLI.md) |\n| Making a trace | [skills\u002Fmake-trace\u002FSKILL.md](skills\u002Fmake-trace\u002FSKILL.md), or run `\u002Fmake-trace` |\n| The format spec | [SCHEMA.md](docs\u002Ftrace\u002FSCHEMA.md) and [FIELDS.md](docs\u002Ftrace\u002FFIELDS.md) |\n| All examples | [docs\u002FEXAMPLES.md](docs\u002FEXAMPLES.md) |\n\n---\n\n## Community\n\n**If Flowtrace is useful to you, consider starring the repo. It helps others find it.**\n\n- **Contributing**: see [CONTRIBUTING.md](.\u002FCONTRIBUTING.md), and look for [good first issues](https:\u002F\u002Fgithub.com\u002FAIScientists-Dev\u002FFlowtrace\u002Flabels\u002Fgood%20first%20issue).\n- **GitHub Issues**: [report bugs \u002F propose changes](https:\u002F\u002Fgithub.com\u002FAIScientists-Dev\u002Fflowtrace\u002Fissues)\n- **Discord**: [discord.gg\u002Fx9mtbMEx](https:\u002F\u002Fdiscord.gg\u002Fx9mtbMEx)\n- **X**: [@morphmind__ai](https:\u002F\u002Fx.com\u002Fmorphmind__ai?s=11)\n\n---\n\nMIT. See [`LICENSE`](.\u002FLICENSE).\n","Flowtrace 是一个将AI执行的任务转化为一系列可保存、重用和优化步骤的工具，而不是一次性聊天。它采用TypeScript编写，支持透明化任务流程，每一步都会在磁盘上生成实际文件供检查，确保结果可追溯且可靠。该工具特别适用于需要结构化处理和重复利用AI工作成果的场景，如自动化决策报告生成等。通过与现有AI代理（例如Claude Code, Codex, Cursor）集成，用户可以轻松构建、修改并跟踪整个任务流程，从而提高工作效率和质量。",2,"2026-06-11 04:10:12","CREATED_QUERY"]