startup-design by ferdinandobons/startup-skill
npx skills add https://github.com/ferdinandobons/startup-skill --skill startup-design一个结构化、多阶段的技能,将创业想法从原始概念转化为经过验证的设计。它生成一套按领域组织的完整 Markdown 文档,并内置进度跟踪功能,确保工作能在会话中断后继续。
该过程按顺序执行 8 个阶段。每个阶段都会生成输出文件并更新进度跟踪器。如果会话中断,则从最后一个完成的检查点恢复。
INTAKE → BRAINSTORM → RESEARCH → STRATEGY → BRAND → PRODUCT → FINANCIAL → VALIDATION
完整模式(默认): 按顺序执行所有 8 个阶段。最适合从头开始彻底设计一个创业项目。
快速通道模式: 当用户表示他们想要"快速验证"、"快速评估"或类似表述,或者当时间/预算明显有限时,运行压缩版本:
快速通道模式生成的文件较少,但仍能为创始人提供基于证据的明确继续/停止信号。在 PROGRESS.md 中注明使用了快速通道模式,以便如果想法通过验证,未来的会话可以扩展到完整模式。
默认输出语言为英语。如果用户使用其他语言书写或明确要求使用某种语言,则所有输出都使用该语言。
参考: 在开始时阅读一次
references/output-guidelines.md。它定义了标准文件头/尾(标题、日期、阶段、置信度、标志)、跨阶段引用格式、优质输出与劣质输出的质量示例,以及如何处理过程中的方向调整。
在进行任何操作之前,检查工作目录(或项目子目录)中是否存在 文件。如果存在,请读取它并从最后一个未完成的阶段恢复。告诉用户:"我发现了之前会话的进度。您已完成 [阶段]。从 [下一个阶段] 继续。"
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
PROGRESS.md如果不存在进度文件,则从阶段 1 开始。
后续所有工作的质量都取决于您现在提取的背景信息的多少。不要仓促进行此阶段——一次彻底的信息收集可以节省后续数小时的错误方向。
以对话式流程提问这些问题,而不是作为僵化的清单。自然地分组相关问题,并根据答案进行调整。并非每个问题都适用于每个创业项目——跳过不相关的内容。
想法
创始人
市场
业务
约束与偏好
在核心问题之后,有意识地提出这些令人不适的问题。它们能及早发现盲点:
不要跳过这些问题——它们为整个过程定下基调,并表明这是一次诚实的评估,而不是一场加油助威会。
将整合后的信息收集内容保存到 {project-name}/00-intake/brief.md,所有捕获的信息都要清晰组织。项目名称应源自创业想法(kebab-case,例如 pet-health-tracker)。
在项目根目录创建 PROGRESS.md,包含:项目名称、开始日期、语言、所有 8 个阶段的清单(标记阶段 1 为完成),以及一个用于会话状态的 Notes 部分。
在深入研究之前,先探索想法空间。这可以防止过早地收敛于想法的第一个版本。
保存到 {project-name}/00-intake/brainstorm.md。更新 PROGRESS.md。
信息收集(以及头脑风暴,如果适用)之后,评估市场复杂性并向用户展示研究深度建议。
参考: 阅读
references/research-scaling.md了解复杂性评分矩阵、层级定义、波次配置和用户沟通模板。
research-scaling.md 获取确切模板)选定的层级决定了阶段 3 中每个波次的代理数量和每个代理的搜索轮次。有关每个层级的精确波次配置,请参见 research-scaling.md。
这是最耗费资源的阶段。它使用 4 个连续的网页研究波次,每个波次都建立在前一个波次的发现之上。
检查 Agent 工具是否可用(Claude Code)或不可用(Claude.ai,其他环境):
阶段 3 需要 WebSearch。在 Claude Code 中,该工具始终可用——如果用户没有预先批准,系统会提示用户进行每次搜索。如果用户拒绝许可,或者在根本不存在 WebSearch 的环境中,则回退到基于知识的研究模式:使用您的训练数据,用 [Knowledge-Based — not live data, verify independently] 清晰标记所有发现,将置信度评级降低一个级别,并建议创始人手动验证关键主张。在 PROGRESS.md 中注明该模式,以便未来的会话知道研究不是基于网页来源的。
参考资料 — 为每个波次阅读相关文件:
references/research-principles.md— 跨领域规则(来源质量、交叉引用、量化、处理搜索失败)。首先阅读此文件。references/research-wave-1-market.md— 第 1 波代理模板(市场规模、趋势、监管)references/research-wave-2-competitors.md— 第 2 波代理模板(直接、间接、GTM 分析)references/research-wave-3-customers.md— 第 3 波代理模板(客户声音、需求、受众)references/research-wave-4-distribution.md— 第 4 波代理模板(渠道、地理市场进入)references/research-synthesis.md— 如何将原始发现综合成最终交付物
仅阅读原则文件 + 您当前正在执行的波次文件。不要一次性加载所有波次文件。
第 1 波:市场概况(3 个代理并行,或 3 个顺序研究块)
在开始第 2 波之前完成第 1 波。将关键发现作为上下文传递。
第 2 波:竞争分析(3 个代理并行,或 3 个顺序研究块)
在开始第 3 波之前完成第 2 波。将竞争对手列表和 GTM 发现作为上下文传递。
第 3 波:客户与需求(3 个代理并行,或 3 个顺序研究块)
在开始第 4 波之前完成第 3 波。
第 4 波:分销与合作伙伴(2 个代理并行,或 2 个顺序研究块)
所有代理将原始发现保存到 {project-name}/01-discovery/raw/。所有波次完成后,综合成 4 个精炼的交付物。综合必须:
{project-name}/01-discovery/market-analysis.md — 市场规模(TAM/SAM/SOM)、增长、成熟度、监管摘要、时机评估{project-name}/01-discovery/competitor-landscape.md — 竞争对手档案、结构化比较矩阵(表格列:名称、产品、定价、目标、融资、进展、关键优势、关键劣势)、定位图、平台风险、脆弱性分析{project-name}/01-discovery/target-audience.md — 用户画像、痛点层次结构、待完成工作、语言地图、购买行为、渠道{project-name}/01-discovery/industry-trends.md — 技术趋势、投资信号、行为转变、监管轨迹、战略含义{project-name}/01-discovery/confidence-dashboard.md — 所有研究数据质量的摘要。对于每个主要主张,列出:主张、来源层级 (1/2/3)、佐证来源数量、置信度级别(高/中/低)、数据时效。这告诉创始人他们在哪些方面基础扎实,哪些方面基础薄弱。更新 PROGRESS.md。
综合完成并所有交付物文件编写完毕后,运行一次验证检查以发现不一致之处。
参考: 阅读
references/verification-agent.md获取完整的验证协议、通用检查和技能特定检查。
01-discovery/ 中的所有交付物文件,并检查:未标记的主张、内部矛盾、置信度评级一致性、缺失的数据缺口、缺失的标志、过时数据以及重复来源的虚假佐证{project-name}/01-discovery/verification-report.md在 Claude.ai 或 Agent 工具不可用时,请遵循相同协议在主对话中自行运行验证检查。
在投入时间进行战略到验证阶段之前,暂停并根据研究发现向创始人展示诚实的评估。这是一个决策点,而不是形式。
呈现简要摘要:"以下是研究发现的内容。" 涵盖市场规模、竞争强度、客户需求信号和时机。然后给出明确的建议:
询问创始人:"基于此,您希望继续进行全面战略、调整想法还是就此停止?" 尊重他们的决定,但要确保这是一个知情的决定。将关口评估保存在 {project-name}/01-discovery/research-gate.md 中。
手握研究结果,定义战略基础。每个文档都应引用阶段 3 的具体发现——脱离研究的战略只是猜测。
参考: 阅读
references/frameworks.md了解精益画布、April Dunford 定位、价值主张画布以及 RICE/MoSCoW 优先级排序的规范定义。使用这些来确保每个框架的一致、准确应用。
在 02-strategy/lean-canvas.md 中构建完整的精益画布(单页商业模式):
在 02-strategy/value-proposition.md 中定义:
在 02-strategy/business-model.md 中详细说明:
在 02-strategy/positioning.md 中,使用 April Dunford 的定位框架:
在 02-strategy/go-to-market.md 中:
更新 PROGRESS.md。
检查点: 开始之前,向创始人简要介绍战略摘要:定位、目标市场、商业模式。询问:"这反映了您的愿景吗?在我们在此基础上建立品牌之前,有什么需要调整的吗?"
将战略转化为品牌标识。品牌应该感觉像是定位的自然延伸——而不是事后才想到的。
在 03-brand/mission-vision-values.md 中:
为用户生成 2-3 个使命和愿景选项以供选择或混合。
在 03-brand/tone-of-voice.md 中:
在 03-brand/brand-personality.md 中:
更新 PROGRESS.md。
定义产品,足以开始构建或向开发团队进行简报。使用 01-discovery/competitor-landscape.md 中的竞争对手功能分析和 01-discovery/target-audience.md 中的客户痛点层次结构来指导功能决策——不要闭门造车。
参考: 使用
references/frameworks.md中的 RICE 或 MoSCoW 进行功能优先级排序。
在 04-product/mvp-definition.md 中:
在 04-product/feature-prioritization.md 中:
在 04-product/user-journey.md 中:
更新 PROGRESS.md。
检查点: 在进行预测之前,与创始人确认关键假设:定价、目标客户数量、团队规模、时间线。这些直接驱动数字——在这里出错意味着预测是虚构的。
将战略落实到数字上。诚实地对待假设——将所有内容标记为估计值并解释推理过程。从 01-discovery/market-analysis.md 中提取单位经济性基准(CAC、LTV、流失率、ACV),从 01-discovery/competitor-landscape.md 中提取竞争对手定价,以便将预测锚定在真实数据上。
参考: 阅读
references/industry-benchmarks.md了解按商业模式类型(SaaS、市场平台、电子商务等)划分的标准指标。将创始人的预测与这些基准进行比较,并标记任何超出正常范围的情况——无论是过于悲观还是过于乐观。
在 05-financial/revenue-model.md 中:
在 05-financial/cost-structure.md 中:
在 05-financial/projections.md 中:
更新 PROGRESS.md。
这是最具可操作性的阶段——它告诉创始人接下来具体要做什么来测试想法是否可行。
在 06-validation/validation-playbook.md 中:
根据具体想法定制实验——B2B SaaS 需要的验证与消费者市场平台不同。
在 06-validation/risk-analysis.md 中:
在 06-validation/assumptions-tracker.md 中:
格式化为表格,便于扫描和更新。
在 06-validation/experiment-design.md 中:
在 06-validation/kill-criteria.md 中,定义 5-7 个具体的、可衡量的条件,在这些条件下创始人应该停止或调整方向。将每个条件与一个验证实验联系起来。要具体:"如果少于 3/10 的访谈对象表示他们会支付 $X",而不是"如果没有需求"。这可以保护创始人免受沉没成本思维的影响。
在验证部分的最后,在 06-validation/scorecard.md 中生成一个摘要记分卡:
| 维度 | 分数 (1-10) | 理由 |
|---|---|---|
| 问题严重性 | ||
| 市场规模 | ||
| 竞争优势 | ||
| 可行性 | ||
| 商业模式清晰度 | ||
| 创始人-市场契合度 | ||
| 时机 | ||
| 总体 |
要诚实。如果想法有弱点,请明确指出。目标是帮助创始人做出好的决定,而不是验证他们的自我。在表格后包含一个清晰的裁决段落,给出明确的建议(参见"绝对诚实协议"中的评分指南)。
更新 PROGRESS.md — 标记所有阶段为完成。
所有阶段完成后,首先在对话中打印最终评估仪表板(参见 references/output-guidelines.md,"Final Assessment Dashboard" 部分)。这为创始人提供了一个即时视觉摘要,涵盖所有关键发现,然后他们再深入研究文件。
然后生成两个最终文件:
项目根目录下的 README.md — 执行摘要:
action-plan-30-days.md — 第一个月的具体每周计划:
反模式检查: 在最终确定之前,扫描整个输出以查找常见的创始人反模式,并标记您发现的任何反模式:"为解决方案寻找问题"、"试图煮沸海洋"(一次性涉及太多功能/市场)、"过早扩展"、"虚荣指标"、"在隐秘模式下构建太久"、"忽视单位经济性"。如果存在任何反模式,请在 README 中包含一个简短的检测到的反模式部分。
参考: 每次会话开始时阅读
references/honesty-protocol.md获取完整协议。关键规则总结如下。
此技能帮助创始人做出好的决定,而不是感觉良好。诚实是不可妥协的:
references/ 目录包含支持文档。仅阅读当前阶段所需的内容。
| 文件 | 何时阅读 | 行数 |
|---|---|---|
output-guidelines.md | 每次会话开始时(一次) | ~78 |
honesty-protocol.md | 每次会话开始时(一次) | ~69 |
research-principles.md | 开始阶段 3 之前(一次) | ~54 |
research-wave-1-market.md | 生成第 1 波代理时 | ~206 |
research-wave-2-competitors.md | 生成第 2 波代理时 | ~220 |
research-wave-3-customers.md | 生成第 3 波代理时 | ~233 |
research-wave-4-distribution.md | 生成第 4 波代理时 | ~132 |
research-synthesis.md | 所有波次完成后,编写最终文件之前 | ~104 |
research-scaling.md | 信息收集之后,阶段 3 之前 | ~95 |
verification-agent.md | 综合之后,阶段 3.5 之前 | ~100 |
frameworks.md | 在阶段 4(战略)、阶段 6(产品)和阶段 8(验证)期间 | ~110 |
industry-benchmarks.md | 在阶段 7(财务)期间 | ~80 |
每周安装量
72
仓库
GitHub 星标数
149
首次出现
Mar 7, 2026
安全审计
安装于
kimi-cli71
github-copilot71
gemini-cli71
amp71
cline71
codex71
A structured, multi-phase skill that takes a startup idea from raw concept to validated design. It produces a complete set of markdown documents organized by domain, with built-in progress tracking so work survives session interruptions.
The process has 8 phases executed sequentially. Each phase produces output files and updates the progress tracker. If a session is interrupted, resume from the last completed checkpoint.
INTAKE → BRAINSTORM → RESEARCH → STRATEGY → BRAND → PRODUCT → FINANCIAL → VALIDATION
Full Mode (default): Execute all 8 phases in order. Best for thoroughly designing a startup from scratch.
Fast Track Mode: When the user says they want a "quick validation," "rapid assessment," or similar, or when time/budget is clearly limited, run a compressed version:
Fast Track produces fewer files but still gives the founder a clear go/no-go signal with evidence. Note in PROGRESS.md that Fast Track mode was used, so a future session can expand to full mode if the idea passes validation.
Default output language is English. If the user writes in another language or explicitly requests one, use that language for all outputs instead.
Reference: Read
references/output-guidelines.mdonce at the start. It defines the standard file header/footer (title, date, phase, confidence, flags), cross-phase referencing format, quality examples of good vs. bad output, and how to handle mid-process pivots.
Before anything else, check if a PROGRESS.md file exists in the working directory (or a project subdirectory). If it does, read it and resume from the last incomplete phase. Tell the user: "I found progress from a previous session. You completed [phases]. Picking up from [next phase]."
If no progress file exists, start from Phase 1.
The quality of everything downstream depends on how much context you extract now. Don't rush this — a thorough intake saves hours of misdirection later.
Ask these in a conversational flow, not as a rigid checklist. Group related questions naturally and adapt based on answers. Not every question applies to every startup — skip what's irrelevant.
The Idea
The Founder(s)
The Market
The Business
Constraints & Preferences
After the core questions, ask these deliberately uncomfortable questions. They surface blind spots early:
Don't skip these — they set the tone for the entire process and signal that this is an honest assessment, not a cheerleading session.
Save the consolidated intake to {project-name}/00-intake/brief.md with all captured information organized clearly. The project name should be derived from the startup idea (kebab-case, e.g., pet-health-tracker).
Create PROGRESS.md at the project root with: project name, start date, language, a checklist of all 8 phases (mark Phase 1 complete), and a Notes section for session state.
Before diving into research, explore the idea space. This prevents premature convergence on the first version of the idea.
Diverge — Generate 5-8 variations of the core idea. Push boundaries:
Analyze — For each variation, note:
Converge — Present the variations to the user. Help them identify which elements resonate. The goal isn't to pick one variation — it's to enrich the original idea with insights from the exploration.
Refine — Based on the user's reactions, crystallize the refined idea. Update the brief if the idea evolved significantly.
Save to {project-name}/00-intake/brainstorm.md. Update PROGRESS.md.
After intake (and brainstorm if applicable), assess market complexity and present the Research Depth recommendation to the user.
Reference: Read
references/research-scaling.mdfor the complexity scoring matrix, tier definitions, wave configurations, and the user communication template.
research-scaling.md for the exact template)The selected tier determines the number of agents per wave and search rounds per agent in Phase 3. See research-scaling.md for exact wave configurations per tier.
This is the most resource-intensive phase. It uses 4 sequential waves of web research, each building on the previous one's findings.
Check if the Agent tool is available (Claude Code) or not (Claude.ai, other environments):
Phase 3 requires WebSearch. In Claude Code, the tool is always available — if the user hasn't pre-approved it, the system will prompt them for each search. If the user denies permission, or in environments where WebSearch doesn't exist at all, fall back to Knowledge-Based Research Mode : use your training data, clearly mark all findings with [Knowledge-Based — not live data, verify independently] , reduce confidence ratings by one level, and recommend the founder verify key claims manually. Note the mode in PROGRESS.md so future sessions know the research wasn't web-sourced.
References — Read the relevant file for each wave:
references/research-principles.md— Cross-cutting rules (source quality, cross-referencing, quantification, handling search failures). Read this FIRST.references/research-wave-1-market.md— Agent templates for Wave 1 (market sizing, trends, regulatory)references/research-wave-2-competitors.md— Agent templates for Wave 2 (direct, indirect, GTM analysis)references/research-wave-3-customers.md— Agent templates for Wave 3 (customer voice, demand, audience)references/research-wave-4-distribution.md— Agent templates for Wave 4 (channels, geographic entry)references/research-synthesis.md— How to synthesize raw findings into final deliverables
Read only the principles file + the wave file you're currently executing. Don't load all wave files at once.
Wave 1: Market Landscape (3 agents in parallel, or 3 sequential research blocks)
Complete Wave 1 before starting Wave 2. Pass key findings as context.
Wave 2: Competitive Analysis (3 agents in parallel, or 3 sequential research blocks)
Complete Wave 2 before starting Wave 3. Pass competitor list and GTM findings as context.
Wave 3: Customer & Demand (3 agents in parallel, or 3 sequential research blocks)
Complete Wave 3 before starting Wave 4.
Wave 4: Distribution & Partnerships (2 agents in parallel, or 2 sequential research blocks)
All agents save raw findings to {project-name}/01-discovery/raw/. After all waves complete, synthesize into 4 polished deliverables. The synthesis must:
{project-name}/01-discovery/market-analysis.md — Market size (TAM/SAM/SOM), growth, maturity, regulatory summary, timing assessment{project-name}/01-discovery/competitor-landscape.md — Competitor profiles, structured comparison matrix (table with columns: Name, Product, Pricing, Target, Funding, Traction, Key Strength, Key Weakness), positioning map, platform risk, vulnerability analysis{project-name}/01-discovery/target-audience.md — Persona(s), pain hierarchy, jobs-to-be-done, language map, buying behavior, channels{project-name}/01-discovery/industry-trends.md — Tech trends, investment signals, behavioral shifts, regulatory trajectory, strategic implications{project-name}/01-discovery/confidence-dashboard.md — Summary of data quality across all research. For each major claim, list: the claim, source tier (1/2/3), number of corroborating sources, confidence level (High/Medium/Low), and data age. This tells the founder where they're standing on solid ground vs. thin ice.Update PROGRESS.md.
After synthesis completes and all deliverable files are written, run a verification pass to catch inconsistencies.
Reference: Read
references/verification-agent.mdfor the full verification protocol, universal checks, and skill-specific checks.
01-discovery/ and checks for: unlabeled claims, internal contradictions, confidence rating consistency, missing data gaps, missing flags, stale data, and duplicate-source false corroboration{project-name}/01-discovery/verification-report.mdIn Claude.ai or when Agent tool is unavailable, run the verification checks yourself in the main conversation following the same protocol.
Before investing time in Strategy through Validation, pause and present the founder with an honest assessment based on research findings. This is a decision point, not a formality.
Present a brief summary: "Here's what the research found." Cover market size, competition intensity, customer demand signals, and timing. Then give a clear recommendation:
Ask the founder: "Based on this, do you want to continue to full strategy, pivot the idea, or stop here?" Respect their decision, but make sure it's an informed one. Save the gate assessment in {project-name}/01-discovery/research-gate.md.
With research in hand, define the strategic foundations. Each document should reference specific findings from Phase 3 — strategy disconnected from research is just guessing.
Reference: Read
references/frameworks.mdfor canonical definitions of Lean Canvas, April Dunford Positioning, Value Proposition Canvas, and RICE/MoSCoW prioritization. Use these to ensure consistent, accurate application of each framework.
Build a complete Lean Canvas (1-page business model) in 02-strategy/lean-canvas.md:
In 02-strategy/value-proposition.md, define:
In 02-strategy/business-model.md, detail:
In 02-strategy/positioning.md, using April Dunford's positioning framework:
In 02-strategy/go-to-market.md:
Update PROGRESS.md.
Checkpoint: Before starting, briefly present the strategy summary to the founder: positioning, target market, business model. Ask: "Does this reflect your vision? Anything to adjust before we build the brand on top of it?"
Translate strategy into brand identity. The brand should feel like a natural extension of the positioning — not an afterthought.
In 03-brand/mission-vision-values.md:
Generate 2-3 options for mission and vision for the user to choose from or remix.
In 03-brand/tone-of-voice.md:
In 03-brand/brand-personality.md:
Update PROGRESS.md.
Define the product enough to start building or to brief a development team. Use the competitor feature analysis from 01-discovery/competitor-landscape.md and customer pain hierarchy from 01-discovery/target-audience.md to inform feature decisions — don't design in a vacuum.
Reference: Use RICE or MoSCoW from
references/frameworks.mdfor feature prioritization.
In 04-product/mvp-definition.md:
In 04-product/feature-prioritization.md:
In 04-product/user-journey.md:
Update PROGRESS.md.
Checkpoint: Before projections, confirm key assumptions with the founder: pricing, target customer volume, team size, timeline. These directly drive the numbers — getting them wrong here means the projections are fiction.
Ground the strategy in numbers. Be honest about assumptions — label everything as estimated and explain the reasoning. Pull unit economics benchmarks (CAC, LTV, churn, ACV) from 01-discovery/market-analysis.md and competitor pricing from 01-discovery/competitor-landscape.md to anchor projections in real data.
Reference: Read
references/industry-benchmarks.mdfor standard metrics by business model type (SaaS, marketplace, e-commerce, etc.). Compare the founder's projections against these benchmarks and flag any that fall outside normal ranges — both too pessimistic and too optimistic.
In 05-financial/revenue-model.md:
In 05-financial/cost-structure.md:
In 05-financial/projections.md:
Update PROGRESS.md.
This is the most actionable phase — it tells the founder exactly what to do next to test whether the idea works.
In 06-validation/validation-playbook.md:
Tailor experiments to the specific idea — a B2B SaaS needs different validation than a consumer marketplace.
In 06-validation/risk-analysis.md:
In 06-validation/assumptions-tracker.md:
Format as a table for easy scanning and updating.
In 06-validation/experiment-design.md:
In 06-validation/kill-criteria.md, define 5-7 specific, measurable conditions under which the founder should stop or pivot. Tie each to a validation experiment. Be specific: "If fewer than 3/10 interview subjects say they'd pay $X" not "if there's no demand." This protects the founder from sunk-cost thinking.
At the end of the validation section, produce a summary scorecard in 06-validation/scorecard.md:
| Dimension | Score (1-10) | Rationale |
|---|---|---|
| Problem severity | ||
| Market size | ||
| Competitive advantage | ||
| Feasibility | ||
| Business model clarity | ||
| Founder-market fit | ||
| Timing | ||
| Overall |
Be honest. If the idea has weaknesses, say so clearly. The goal is to help the founder make a good decision, not to validate their ego. Include a clear Verdict paragraph after the table with an unambiguous recommendation (see the scoring guide in the Radical Honesty Protocol).
Update PROGRESS.md — mark all phases complete.
After all phases are complete, first print the Final Assessment Dashboard in the conversation (see references/output-guidelines.md, "Final Assessment Dashboard" section). This gives the founder an instant visual summary of all key findings before they dive into the files.
Then produce two final files:
README.md at the project root — executive summary:
action-plan-30-days.md — concrete weekly plan for the first month:
Anti-pattern check: Before finalizing, scan the entire output for common founder anti-patterns and flag any you detect: "solution looking for a problem," "boiling the ocean" (too many features/markets at once), "premature scaling," "vanity metrics," "building in stealth too long," "ignoring unit economics." Include a brief Anti-Patterns Detected section in the README if any are present.
Reference: Read
references/honesty-protocol.mdat the start of every session for the full protocol. The key rules are summarized here.
This skill helps founders make good decisions, not feel good. Honesty is non-negotiable:
The references/ directory contains supporting documentation. Read only what you need for the current phase.
| File | When to Read | Lines |
|---|---|---|
output-guidelines.md | At the start of every session (once) | ~78 |
honesty-protocol.md | At the start of every session (once) | ~69 |
research-principles.md | Before starting Phase 3 (once) | ~54 |
research-wave-1-market.md | When spawning Wave 1 agents | ~206 |
research-wave-2-competitors.md |
Weekly Installs
72
Repository
GitHub Stars
149
First Seen
Mar 7, 2026
Security Audits
Gen Agent Trust HubPassSocketPassSnykWarn
Installed on
kimi-cli71
github-copilot71
gemini-cli71
amp71
cline71
codex71
任务估算指南:敏捷开发故事点、计划扑克、T恤尺码法详解
10,500 周安装
| When spawning Wave 2 agents |
| ~220 |
research-wave-3-customers.md | When spawning Wave 3 agents | ~233 |
research-wave-4-distribution.md | When spawning Wave 4 agents | ~132 |
research-synthesis.md | After all waves complete, before writing final files | ~104 |
research-scaling.md | After intake, before Phase 3 | ~95 |
verification-agent.md | After synthesis, before Phase 3.5 | ~100 |
frameworks.md | During Phase 4 (Strategy), Phase 6 (Product), and Phase 8 (Validation) | ~110 |
industry-benchmarks.md | During Phase 7 (Financial) | ~80 |