[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"pack-detail-spec-driven-ai-dev-en":3,"seo:pack:spec-driven-ai-dev:en":62},{"code":4,"message":5,"data":6},200,"操作成功",{"pack":7},{"slug":8,"icon":9,"tone":10,"status":11,"status_label":12,"title":13,"description":14,"items":15,"install_cmd":61},"spec-driven-ai-dev","📝","#404040","stable","Stable","Spec-Driven AI Dev","OpenSpec, Planning with Files, OpenDeepWiki, the GSD meta-prompt system, Rivet visual IDE — write the spec first, let the agent execute.",[16,28,35,45,52],{"id":17,"uuid":18,"slug":19,"title":20,"description":21,"author_name":22,"view_count":23,"vote_count":24,"lang_type":25,"type":26,"type_label":27},340,"5720ef91-76d7-4586-8fe8-71ba8c921d58","openspec-spec-driven-ai-development-5720ef91","OpenSpec — Spec-Driven AI Development","OpenSpec provides structured specifications that AI coding agents follow to produce consistent code. 36K+ stars. Works with Cursor, Claude Code, Copilot. MIT.","TokRepo精选",333,0,"en","skill","Skill",{"id":29,"uuid":30,"slug":31,"title":32,"description":33,"author_name":34,"view_count":23,"vote_count":24,"lang_type":25,"type":26,"type_label":27},472,"034be597-c361-45a2-b143-41cce0ec8ad8","planning-files-manus-style-persistent-planning-skill-034be597","Planning with Files — Manus-Style Persistent Planning Skill","Claude Code skill implementing persistent markdown planning with 96.7% benchmark pass rate. Uses a 3-file pattern (task_plan.md, findings.md, progress.md) to survive context resets.","Skill Factory",{"id":36,"uuid":37,"slug":38,"title":39,"description":40,"author_name":41,"view_count":42,"vote_count":24,"lang_type":25,"type":43,"type_label":44},503,"24613482-86e1-487a-a2ad-450c9bb7dfaa","opendeepwiki-turn-any-repo-into-ai-documentation-24613482","OpenDeepWiki — Turn Any Repo into AI Documentation","Self-hosted tool that converts GitHub, GitLab, and Gitea repositories into AI-powered knowledge bases with Mermaid diagrams and conversational AI. MIT license, 3,000+ stars.","Script Depot",401,"script","Script",{"id":46,"uuid":47,"slug":48,"title":49,"description":50,"author_name":51,"view_count":29,"vote_count":24,"lang_type":25,"type":26,"type_label":27},712,"e108cf5c-c34e-4d27-a694-66a693301e87","get-shit-done-gsd-meta-prompting-dev-system-claude-code-e108cf5c","Get Shit Done (GSD) — Meta-Prompting Dev System for Claude Code","A spec-driven development system with 48.6k GitHub stars. Adds phase-based planning, multi-agent execution, verification gates, and state persistence to Claude Code, Cursor, Gemini CLI and 9 more runtimes. Install with one npx command.","henuwangkai",{"id":53,"uuid":54,"slug":55,"title":56,"description":57,"author_name":41,"view_count":58,"vote_count":24,"lang_type":25,"type":59,"type_label":60},749,"9202a5c4-0998-449b-beb2-9dfcce135e37","rivet-visual-ai-prompt-workflow-ide-9202a5c4","Rivet — Visual AI Prompt Workflow IDE","Visual IDE for designing and debugging AI prompt chains. Drag-and-drop nodes for LLM calls, conditionals, loops, and data transforms with real-time execution preview.",336,"prompt","Prompt","tokrepo install pack\u002Fspec-driven-ai-dev",{"pageType":63,"pageKey":8,"locale":25,"title":64,"metaDescription":65,"h1":13,"tldr":66,"bodyMarkdown":67,"faq":68,"schema":84,"internalLinks":93,"citations":106,"wordCount":119,"generatedAt":120},"pack","Spec-Driven AI Dev: 5 tools for spec-first agent workflows","OpenSpec, Planning with Files, OpenDeepWiki, GSD meta-prompt, and Rivet visual IDE — write the spec first, let the agent execute. TokRepo install.","Five tools for the spec-driven workflow: write a falsifiable spec, let the AI agent execute against it, and use the spec as the audit trail. Replaces vibe-coding with reviewable artifacts.","## What's in this pack\n\nThis pack collects the **five spec-driven tools** that make AI coding auditable. Each one tackles a different part of the spec → agent → review loop, and they compose well in a single project.\n\n| # | Tool | Layer | What it does |\n|---|---|---|---|\n| 1 | OpenSpec | spec format | Markdown spec with deltas; agent reads & updates it |\n| 2 | Planning with Files | planning loop | Forces agents to write a plan file before editing code |\n| 3 | OpenDeepWiki | knowledge base | Auto-generates a wiki from your repo; spec context |\n| 4 | GSD meta-prompt | prompt system | Convention for breaking a spec into phases \u002F plans |\n| 5 | Rivet | visual IDE | Graph editor for multi-step prompt \u002F agent flows |\n\nOpenSpec and Planning with Files are the load-bearing pair. OpenDeepWiki gives the agent the codebase context it needs to write a competent spec, and Rivet is the visual layer for teams that prefer node graphs over markdown.\n\n## Why spec-driven\n\n\"Vibe coding\" — telling the agent what you want and hoping for the best — works for tiny tasks and falls apart on anything multi-day. Three problems compound:\n\n- **Drift.** The agent loses track of intent across many turns. Each new turn re-derives goals from chat history, which is lossy.\n- **Audit.** When the PR lands you cannot tell what the agent thought it was building. Reviewers reverse-engineer intent from code diffs.\n- **Branching.** Three concurrent attempts at the same feature have no shared source of truth.\n\nA spec fixes all three. Intent lives in a markdown file the agent reads on every turn (no drift). The diff between specs versions is the audit trail (no reverse-engineering). Concurrent attempts fork the spec the same way they fork code (clean branching).\n\n## Install in one command\n\n```bash\n# Install the entire pack\ntokrepo install pack\u002Fspec-driven-ai-dev\n\n# Or pick the core pair\ntokrepo install openspec\ntokrepo install planning-with-files\n```\n\nOpenSpec lands as a `specs\u002F` directory plus a Claude Code subagent that reads and updates it. Planning with Files installs a hook that prompts the agent to emit a `PLAN.md` before editing any code. OpenDeepWiki runs as a separate process that crawls your repo and serves a queryable wiki. Rivet is a desktop app and is installed via `npm install -g @ironclad\u002Frivet`.\n\n## Common pitfalls\n\n- **Specs that are too coarse.** \"Build a checkout flow\" is a goal, not a spec. A usable spec lists acceptance criteria, edge cases, and out-of-scope items. OpenSpec's template enforces this structure; resist the urge to delete the sections you find awkward.\n- **Specs that are too fine.** A 500-line spec for a 20-line patch is over-engineering. Match spec depth to task risk: tiny patches get a one-paragraph spec, multi-week features get the full template.\n- **Spec rot.** When the implementation diverges from the spec, the spec becomes a lie. The agent will follow the lie. Keep the spec live: every PR updates it or explicitly notes the deviation.\n- **Confusing planning files with specs.** PLAN.md is the agent's reasoning trace for the *next batch* of work; the spec is the durable artifact. Don't conflate them; the planning file is throwaway.\n- **Rivet for everything.** Rivet shines for branching prompt flows; it is overkill for single-shot coding tasks. Use it when the graph adds clarity, not as a default.\n\n## Relationship to other packs\n\n- **Anthropic Builders** ships the agent runtime; this pack adds the spec layer on top.\n- **Prompt Engineering Toolkit** sharpens the *language* used inside specs and plans.\n- **LLM Eval & Guardrails** lets you assert that the agent's output matches the spec's acceptance criteria.\n\n## When this pack alone isn't enough\n\nSpec-driven dev assumes you already know what you want. For exploratory work — \"what should this product even be?\" — start with a discovery process (interviews, prototypes, sketches), then translate findings into specs. Skipping that step yields beautifully-formatted specs for the wrong product. The pack is a force multiplier for clarity, not a substitute for it.\n\nA second limit: specs do not replace tests. The spec describes intent and acceptance criteria in prose; tests verify those criteria mechanically. Most teams keep both — the spec for the human reviewer, the test for the CI machine — and let the agent author both during the same turn.",[69,72,75,78,81],{"q":70,"a":71},"Is the pack free?","Yes. OpenSpec, Planning with Files, OpenDeepWiki, the GSD meta-prompt system, and Rivet are all open source. You only pay for the LLM API calls when running the agent against a spec, billed by your provider. The TokRepo install does not introduce a proxy or token. Audit-friendly for security and procurement reviews.",{"q":73,"a":74},"How does this compare to writing prompts in CLAUDE.md?","CLAUDE.md is project-level conventions (style, libraries to prefer). Specs are feature-level intent (what we're building, why, how to know it's done). They complement, not replace. A mature project has a stable CLAUDE.md and a per-feature spec under specs\u002F. The agent reads both on every turn — CLAUDE.md tells it how to write code; the spec tells it what to write.",{"q":76,"a":77},"Works with Claude Code or Cursor?","OpenSpec ships with Claude Code subagent integration first; the spec format itself is plain markdown that any AI editor can read. Planning with Files is also Claude-Code-native via hooks but the planning convention transfers manually to Cursor or Codex CLI. OpenDeepWiki and Rivet are agent-agnostic — they expose HTTP endpoints. The GSD meta-prompt is markdown.",{"q":79,"a":80},"Diff vs writing a PRD before coding?","A PRD is for humans planning a quarter ahead. A spec in this pack is for an agent reading it on this turn. They overlap (both state intent) but differ in granularity, audience, and lifecycle. PRDs are quarterly, owned by PM. Specs are per-feature, owned by the engineer who will implement, and *updated* during implementation rather than frozen at kickoff.",{"q":82,"a":83},"Operational gotcha?","Spec rot is the silent killer. The first month feels great. Then someone makes a quick fix without updating the spec, the agent reads the stale spec next turn and re-introduces the bug. Add a CI check that flags PRs touching code paths whose specs haven't been updated in N days. OpenSpec ships such a check; turn it on.",{"@context":85,"@type":86,"name":13,"description":87,"numberOfItems":88,"publisher":89},"https:\u002F\u002Fschema.org","CollectionPage","OpenSpec, Planning with Files, OpenDeepWiki, GSD meta-prompt and Rivet — spec-first workflows for AI coding agents.",5,{"@type":90,"name":91,"url":92},"Organization","TokRepo","https:\u002F\u002Ftokrepo.com",[94,98,102],{"url":95,"anchor":96,"reason":97},"\u002Fen\u002Fpacks\u002Fanthropic-builders","Anthropic Builders","the official skills the spec drives",{"url":99,"anchor":100,"reason":101},"\u002Fen\u002Fpacks\u002Fprompt-engineering-toolkit","Prompt Engineering Toolkit","complementary discipline for the spec text itself",{"url":103,"anchor":104,"reason":105},"\u002Fen\u002Ftools\u002Fclaude-code","Claude Code","the agent that executes against the spec",[107,111,115],{"claim":108,"source_name":109,"source_url":110},"OpenSpec is an open-source spec-driven workflow for Claude Code","Fission-AI\u002FOpenSpec","https:\u002F\u002Fgithub.com\u002FFission-AI\u002FOpenSpec",{"claim":112,"source_name":113,"source_url":114},"Spec-driven development is GitHub's recommended pattern for agent coding","GitHub blog: spec-driven AI development","https:\u002F\u002Fgithub.blog\u002Fai-and-ml\u002Fgenerative-ai\u002Fspec-driven-development-with-ai-get-started-with-a-new-open-source-toolkit\u002F",{"claim":116,"source_name":117,"source_url":118},"Rivet is a visual graph IDE for prompt and agent flows","Ironclad\u002Frivet","https:\u002F\u002Fgithub.com\u002FIronclad\u002Frivet",736,"2026-05-02T15:30:00Z"]