autonomous-requirement-discovery
Skill263 starsChanged 39 days agoMITView on GitHub
仅当用户明确要求 AI 从产品方向、愿景、真实现状或用户信号中主动发现值得解决的问题、形成候选需求,或降低对详细需求输入的依赖时使用;负责形成有证据、可验证的需求假设,不负责实现、排期或自动纳入路线图。
In this repository37 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .cursor/rules/code-citation-clickable.mdcCursor rule
- .cursor/rules/commit-message-english.mdcCursor rule
- autonomous-maintainability-campaignSkill
- delivering-delegated-linear-issuesSkill
- development-deliverySkill
- development-designSkill
- development-implementationSkill
- development-lifecycleSkill
- development-retrospectiveSkill
- development-reviewSkill
- development-task-understandingSkill
- development-validationSkill
- idea-search-funnelSkill
- nextclaw-agent-instructions-governanceSkill
- nextclaw-product-blog-storytellingSkill
- nextclaw-product-visual-assetsSkill
- x-twitter-birdSkill
- aigen-image-generationSkill
- bb-browserSkill
- browser-controlSkill
- claude-code-narp-runtimeSkill
- clawhub-nextclawSkill
- codex-narp-runtimeSkill
- find-skillsSkill
- gemigo-cliSkill
- impeccableSkill
- lark-cliSkill
- linear-cliSkill
- mmx-cliSkill
- nextclaw-skin-studioSkill
- opencliSkill
- proxy-local-ai-subscriptionsSkill
- publish-to-nextclaw-marketplaceSkill
- skillhub-guideSkill
- superpowersSkill
- ui-ux-pro-maxSkill
The file
Files
自主需求发现
定位
探索“在用户没有给出详细需求时,AI 能否主动发现真正值得解决的问题”。这是独立的探索性能力,不是标准开发生命周期阶段,也不自动接管普通任务理解。
输入可以只有产品方向、目标用户、问题域或上位目标,但必须有明确的探索边界。没有证据时提出可验证假设,不把模型直觉包装成用户需求。
发现循环
- 冻结上位目标、目标用户、探索范围、禁止边界和本轮需要改变的决策。
- 对齐产品愿景,调查当前用户任务、真实产品状态、已有反馈、issue、使用信号和历史决策。
- 区分已确认问题、弱信号、合理假设和纯猜测;缺少用户数据时明确证据等级。
- 为每个高价值候选写清受影响用户、真实阻力、期望结果、支持证据和最小反证方式。
- 用最低成本的只读调查、对照、原型或实验消除最关键不确定性;需要修改产品或外部状态时先取得相应授权。
- 按用户价值、愿景一致性、证据强度、可验证性和实现代价比较候选,不用伪精确总分掩盖判断。
- 推荐一个最值得继续的需求假设;只有真实取舍存在时保留一至两个备选。
需求质量门
候选需求必须同时说明:
- 它解决的是哪一个用户问题,而不是先选定了哪个功能;
- 为什么现在值得解决,证据和未知项分别是什么;
- 用户或系统出现什么可观察变化才算成功;
- 最小可行验证是什么,什么信号会证伪它;
- 哪些相邻问题和解决方案不属于当前范围。
只有“做什么功能”的想法、没有用户结果或反证方式时,继续调查或降级为待验证假设,不进入实现。
输出与停止
输出推荐需求假设、证据、目标用户、成功信号、最小实验、主要未知项、备选与非目标。证据足以支持下一步决策,或继续探索的成本已经高于信息增益时停止。
本 skill 不修改产品实现,不自动创建 roadmap、计划或 issue,也不把探索结论宣称为已确认需求。后续是否进入正式开发由用户或当前任务 owner 决定。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

