Update gr-competitor to 0.4.0 with evidence-led workflows

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Iris / Gingiris
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,30 @@
+---
+license: mit
+language:
+- en
+- zh
+tags:
+- agent-skill
+- gingiris
+- workflow
+pretty_name: gr-competitor — Evidence-led workflow
+---
+# gr-competitor
+
+Version 0.4.0 · Updated 2026-11-08.
+
+适用于分析对手、找对标、竞品扫描与早期付费客户研究；先明确决策与适配条件， 再按证据分析历史、定价、博客和渠道，输出用户假设、需求验收与最小增长实验。
+
+## Install / 安装
+
+```bash
+npx skills add Gingiris-1031/gingiris-skills --skill gr-competitor
+```
+
+Read [SKILL.md](SKILL.md) and its linked references for the complete workflow.
+完整方法与模板请阅读 SKILL.md 及其引用文件。
+
+This update separates observations, hypotheses and outcomes. It contains reusable methods and constructed examples, not private workshop recordings or student records. No guaranteed business outcome.
+本次更新区分观察、假设与实际结果；仅包含通用方法与教学示例，不含私人逐字稿、学员记录，不保证商业效果。
+
+Source: [maintained GitHub package](https://github.com/Gingiris-1031/gingiris-skills/tree/main/skills/gr-competitor).

SKILL.md

@@ -0,0 +1,21 @@
+---
+name: gr-competitor
+description: '适用于分析对手、找对标、竞品扫描与早期付费客户研究；先明确决策与适配条件， 再按证据分析历史、定价、博客和渠道，输出用户假设、需求验收与最小增长实验。
+
+  '
+metadata:
+  author: Iris / Gingiris
+  version: 0.4.0
+---
+
+# gr-competitor — 研究入口
+
+先记录我方任务、付费者、商业模式、阶段、市场、要做的决定和时间/预算。缺竞品先发现，缺商业模式保留不同假设。轻量研究聚焦 1–2 个对象；不强行批量开页。
+
+完整方法与可填写表格已包含在本包：[研究流程](references/research-workflow.md)。按问题读取其中的本地链接；单独安装也可使用。该流程由 competitor-research-playbook 自动生成，维护时不手改。
+
+当问题是“谁会付费”或“如何把吐槽变需求”，读取 [早期客户证据与需求交接](references/early-customer-evidence.md)。付费身份、早期状态与案例时间分开验证。官网 Logo 不等于付款，用户抱怨不等于已验证解法。
+
+搜索、提取、截图、存档和浏览器按需要选择，actionbook 可用时是一个选项，不是必装依赖，也不承诺固定速度/并发收益。
+
+交付选择理由、带日期的来源、未知项、反证与我方行动卡。用户访谈、B2B、SEO 等后续流程接收该卡；不因发现问题自动发文、联系用户、改价或创建监控任务。

references/early-customer-evidence.md

@@ -0,0 +1,19 @@
+# Early customer evidence → ICP hypotheses
+
+Use when “power users” or a broad product category fails to explain who buys and why. Start with a small set of relevant competitors; prefer dated first-party customer stories and archived early pages. A target such as the first ten paying customers is a research question, not a promise that those records exist.
+
+Record:
+
+`Source URL | Publication / archive date | Evidence of early status | Evidence of payment | Industry | User role / buyer | Job and trigger | Previous workaround | Outcome stated by source | Unknowns`
+
+A logo, testimonial or case study alone does not establish payment or early-customer status. Label those unknown explicitly. Keep source claims, customer statements and your inference separate. Do not infer a full customer list from a few public examples or collect private identities to fill gaps.
+
+Group observed jobs into one to three industry hypotheses, with disconfirming evidence and questions for interviews. “Consultants preparing a client deliverable from scattered notes” is a testable job; “AI power users” alone is not. Competitor evidence can guide recruitment but cannot validate your own PMF.
+
+## Handoff from complaint to a testable change
+
+A complaint describes a problem, not the solution. When product recommendations are requested, add:
+
+`Evidence | User task | Current failure | Proposed solution | Reference implementation / benchmark | Category | Priority rationale | Owner | Acceptance steps | User to revisit`
+
+Categories may be core capability, onboarding or commercial readiness. Select P0/P1/P2 from impact and dependencies; do not label every complaint P0. Describe the starting state, user action and observable result for acceptance. Mark a solution as proposed until tested. If interviews are available, use gr-user-interview to validate; otherwise include the unanswered interview questions in the handoff.

references/research-workflow.md

@@ -0,0 +1,39 @@
+# Competitor Research Playbook / 竞品研究
+
+唯一维护源：`skills/competitor-research-playbook/`。`gr-competitor-research` 是自动生成的兼容名称，带完整本地 references，可独立安装。不要分别编辑；维护者运行 `python3 scripts/sync-competitor.py`，用 `--check` 检测漂移。版本相同意味着正文与方法文件相同（入口名称除外）。
+
+## 任务契约 / Research brief
+
+先明确我方产品、用户任务、付费者、商业模式、阶段、市场、要做的决定及时间/预算上限。缺竞品先发现；缺商业模式则分开记录几个可区分假设。轻量研究选择 1–2 个对象和足够支持决策的窗口，深度研究按证据扩展，不强行凑三家或十家。
+
+开始前写调研停止条件：哪些未知影响决定、可投入多久、什么证据足够设计一次低成本可逆测试。达到条件就交付行动，不无限补表；关键证据缺失仍标未知。
+
+## 执行 / Workflow
+
+1. **选对标。** 按用户/付费者、任务、商业模式、阶段、市场和产品形态，区分直接竞品、渠道参考、定位参考与不适用对象。知名度不代表适配度；不同商业模式可以借鉴某个渠道，不能据此证明我方增长路径。
+2. **还原阶段。** 比较最早可得版本、Beta/发布与当前页面，记录时间和原始链接。最早存档不等于第一版；Beta、发布、PMF 不等价，疑似收敛只标候选转折。读取 [历史与形态分支](research/history.md)。
+3. **查客户证据。** 案例日期、早期身份与实际付款分别举证；Logo、员工邮箱、免费用户不证明企业采购。使用 [客户证据与需求交接](research/early-customer-evidence.md)。增加“是否改变我方待访谈人群/任务”；无法验证付费时不补猜名单。
+4. **拆内容生产系统。** 按观察窗口与样本方法分析栏目、主题比例、频次、内容结构、作者线索和 CTA。署名不等于雇佣关系，产量不等于成本，没收入和完整成本不能称 ROI。
+5. **分渠道找参考。** 官方、创始人、生态、创作者分别记录；历史发布窗口与近期常态分开。YouTube 中插/整条合作区分，未披露赞助只称提及。传播链路记录时间与链接，不凭转发顺序断言付费或因果。近期可见帖子也不代表平台允许我方复发。
+6. **找伙伴。** 从公开 Partner 页区分经销、实施、集成、推荐及未知，保留地区、人群、层级和公开联系入口。这里只形成线索，研究授权不包含外发邀约。
+7. **回到我方定位与测试。** 从用户现有工具/人工流程出发，使用熟悉词汇和可证明差异；同版本、同条件比较，缺证据不打叉。每条建议连接到我方能力/资源与最小测试。
+
+按需填写 [研究表与行动卡](research/research-templates.md)，不为填满模板制造数据。
+
+## 证据与交付 / Evidence and delivery
+
+每项结论记录来源、观察时间、统计窗口、对象范围、事实/公司自报/工具估算/分析假设。观看量、访问、注册、激活、付款不是同一指标；全站规模不能直接归给某条产品线。定价比较写币种、计费单位、周期、税费、席位和版本。工具失效、访问受限、缺存档只说明未知。
+
+报告必须包含：选择理由、证据表、关键未知、我方行动卡以及反证。反证栏回答“竞品真实优势是什么、哪些水下服务能力不可见、什么证据会推翻我方判断”。不能为支持自己的品牌叙事贬低竞品。
+
+行动卡包含最小测试、预算/时间上限、负责人、指标分母/窗口、继续/修改/暂停条件。流程诊断或六阶段飞轮评分仅是整理假设的工具，评分不证明 PMF 或因果；内容形式的优劣必须按场景验证，不能宣称固定排行。
+
+工具按可得权限选择搜索、网页提取、存档、浏览器或现有连接器。截图与结构化文字互补，无强制商业工具依赖；付费数据、安装新软件、外发联系和定时监控按用户授权另行处理。
+
+## English operating summary
+
+Select benchmarks by buyer, job, business model, stage and market. Distinguish direct competitors from channel or positioning references. Separate historical reconstruction from current execution conditions. Date every observation; label company claims, estimates and hypotheses. Logos do not prove early paid customers. End with a bounded experiment and disconfirming evidence. The linked Chinese worksheets are the authoritative operational detail; legacy translations are not maintained as separate methods.
+
+## Decision and evidence review / 决策验收
+
+For page-purpose branches, social metric denominators, misleading growth inferences and a one-page action review, read [decision review](research/decision-review.md). 用它检查结论是否能解释、适用并验证，而不只检查报告是否完整。

references/research/decision-review.md

@@ -0,0 +1,45 @@
+# Decision review / 从研究到决定
+
+Use when a teardown must inform a product, positioning or channel decision. The examples below are constructed teaching examples, not customer results.
+
+## Match the page to the visitor's task / 先识别页面任务
+
+| Page | Visitor's question | Inspect | Candidate outcome |
+|---|---|---|---|
+| Marketing landing page / 营销官网 | Is this for me, why trust it? | Target user, task, evidence, eligibility, next step | Qualified trial or demo |
+| Content or project community / 素材社区 | Can I find and use the right project? | Search, previews, usable files, requirements, first-use path | Relevant project used or completed |
+| Tool workspace / 工具工作台 | How do I finish my task? | Inputs, permissions, defaults, error recovery | Task completion without assistance |
+
+Record acquisition source, existing product knowledge and first value action. Do not compare screen counts across different page purposes or require every community to become a marketing landing page. For a two-sided community, separate creators/suppliers from consumers; existing hardware owners may have different entry needs. A reported conversion drop after a redesign is a hypothesis until cohort, traffic mix and instrumentation are checked.
+
+For products with eligibility constraints, compare supported region/entity, goods or task type, onboarding/approval, buyer, implementation and service dependencies. Mark unsupported eligibility or payment claims unknown until verified from an authoritative source. Price and feature lists alone cannot establish usability.
+
+## Social evidence worksheet / 社媒数据口径
+
+For each dataset record date/timezone, historical vs current stage, account role, sampling/full enumeration, window, missing pages, original/quote/repost/reply inclusion and collection timestamp.
+
+Keep interaction numerator and denominator explicit: (named reactions + comments + shares) / impressions is different from / followers. Current followers are not historical followers. Do not mix datasets with different inclusion rules; show sample size and median alongside a mean when outliers dominate. Public views, site visits, unique visitors, registrations and payments are distinct measures. Separate known paid/organic data; unknown promotion is not zero paid spend.
+
+| Observation | Unsupported leap | Useful next check |
+|---|---|---|
+| Direct is the largest traffic category | Word of mouth caused growth | Attribution gaps, returning users, campaign and referral evidence |
+| One channel paused posting | The whole company stopped marketing | Other channels and creator collaborations in the same window |
+| Several posts share a hook | They were all paid or fake | Disclosure, account roles and corroborating evidence; otherwise unknown |
+| Large views, few interactions | Bots or wasted budget | Distribution and denominator, qualified visits and downstream behavior |
+| A launch and traffic peak coincide | Launch caused all growth | Dates, other events, source-level traffic; retain alternative explanations |
+
+Comments enter the needs worksheet when they describe a task, obstacle, alternative or concrete usage. Generic praise or restated copy is low-information, not proof of a bot. A complaint does not establish willingness to switch or pay.
+
+## One-page decision card / 一页决策卡
+
+- Decision and relevant stage:
+- Observation + original URL/date + sampling limit:
+- Interpretation and plausible alternative explanation:
+- Why it applies to our user/task; why it may not:
+- Proposed change, owner and smallest test:
+- Time/cash cap, metric denominator/window, continue/change/stop condition:
+- What remains unverified:
+
+Before accepting an AI-generated conclusion, answer: where is the source; why does it change our decision; what would overturn it? If the author cannot explain a claim, retrieve evidence, narrow it or mark it unknown. Fluent presentation is not verification. Do not infer a cold-start plan solely from a mature company's current channels.
+
+Constructed example: a file-library homepage emphasizes company positioning while visitors seek a compatible template. Propose a project-first entry for that cohort; measure successful file use as well as signup. Do not promise a conversion lift or assume all traffic has the same intent.

references/research/early-customer-evidence.md

@@ -0,0 +1,19 @@
+# Early customer evidence → ICP hypotheses
+
+Use when “power users” or a broad product category fails to explain who buys and why. Start with a small set of relevant competitors; prefer dated first-party customer stories and archived early pages. A target such as the first ten paying customers is a research question, not a promise that those records exist.
+
+Record:
+
+`Source URL | Publication / archive date | Evidence of early status | Evidence of payment | Industry | User role / buyer | Job and trigger | Previous workaround | Outcome stated by source | Unknowns`
+
+A logo, testimonial or case study alone does not establish payment or early-customer status. Label those unknown explicitly. Keep source claims, customer statements and your inference separate. Do not infer a full customer list from a few public examples or collect private identities to fill gaps.
+
+Group observed jobs into one to three industry hypotheses, with disconfirming evidence and questions for interviews. “Consultants preparing a client deliverable from scattered notes” is a testable job; “AI power users” alone is not. Competitor evidence can guide recruitment but cannot validate your own PMF.
+
+## Handoff from complaint to a testable change
+
+A complaint describes a problem, not the solution. When product recommendations are requested, add:
+
+`Evidence | User task | Current failure | Proposed solution | Reference implementation / benchmark | Category | Priority rationale | Owner | Acceptance steps | User to revisit`
+
+Categories may be core capability, onboarding or commercial readiness. Select P0/P1/P2 from impact and dependencies; do not label every complaint P0. Describe the starting state, user action and observable result for acceptance. Mark a solution as proposed until tested. If interviews are available, use gr-user-interview to validate; otherwise include the unanswered interview questions in the handoff.

references/research/en/README.md

@@ -0,0 +1,3 @@
+# Compatibility language entry
+
+The previous translation is superseded by [version 1.3.0](../../research-workflow.md). The maintained workflow and worksheets are Chinese, with an English operating summary. Do not apply the old independent method.

references/research/history.md

@@ -0,0 +1,14 @@
+# 历史与产品形态分支
+
+先确定问题需要哪个时间窗，不把所有研究固定为近三个月或三年前。
+
+- Web：存档页面、发布公告、定价和早期案例。缺图或脚本失败不证明当年没有功能。
+- App：商店描述、版本记录、开发者公告和评价时间。官网存档缺失时转向这些证据，不编造网站历史。
+- 开源：提交、release、公开公告、Star 变化与当时帖子交叉核对；Star 突增只能定位候选事件，不能单独证明首次开源日期。
+- 硬件：众筹、交付、前代产品、评测与售后证据分开；众筹金额不是常态收入或利润。
+
+历史检索流程：时间限定搜索 → 日期可核验的原始页面/公告 → 该日或该周渠道记录 → 观察与解释分栏。搜索结果日期可能不准确，回到来源校验。融资新闻可形成注意力假设，但不自动意味着现在该发布。
+
+逐字稿中的第三方产品融资额、早期借款、发布效果和 tagline 只作为核验线索，不作为预填事实。未验证的口述案例不进入用户报告。既有来源摘记见 [podcast-evidence.md](podcast-evidence.md)，它不是公开事实核验凭证。
+
+“替代品”定位从用户理解出发，以明确任务和可证明差异表达。历史上的成功帖子、目录或打法不代表当前规则许可、渠道效果或可复现收益。

references/research/ja/README.md

@@ -0,0 +1,3 @@
+# Compatibility language entry
+
+The previous translation is superseded by [version 1.3.0](../../research-workflow.md). The maintained workflow and worksheets are Chinese, with an English operating summary. Do not apply the old independent method.

references/research/ko/README.md

@@ -0,0 +1,3 @@
+# Compatibility language entry
+
+The previous translation is superseded by [version 1.3.0](../../research-workflow.md). The maintained workflow and worksheets are Chinese, with an English operating summary. Do not apply the old independent method.

references/research/podcast-evidence.md

@@ -0,0 +1,7 @@
+# Podcast evidence for competitor research
+
+1. Use star-history/OSS Insight to locate a launch inflection date.
+2. Restrict search to that date/week and reconstruct PR, directories, communities and creator amplification.
+3. Compare first site, beta, first launch and current site.
+4. Benchmark the strongest operator per channel, not one competitor everywhere.
+5. Treat a competitor funding announcement as a category-attention hypothesis, not proof that copying will work.

references/research/research-templates.md

@@ -0,0 +1,69 @@
+# 研究表与行动卡
+
+只填写当前决定需要的表格，空缺写未知并说明影响。
+
+## 1. 对标选择表
+
+| 对象 | 用户/付费者 | 任务/替代流程 | 商业模式 | 阶段与市场 | 直接竞品/渠道参考/定位参考 | 选用理由 | 不可迁移条件 | 来源 |
+|---|---|---|---|---|---|---|---|---|
+
+商业模式相近不是机械排除条件：不同模式也可以学某个渠道或页面，但不能拿它证明我方同一增长路径。不要因为名字最响就优先；不强行凑三个对象。
+
+## 2. 官网与产品阶段表
+
+| 对象 | 存档/发布日期 | 版本与原始链接 | 价值主张 | 面向谁 | 主要任务/CTA | 页面结构与截图 | 定价/计费单位 | 客户案例 | 社区/渠道/伙伴入口 | 与前版变化 | 已观察事实 | 待验证解释 |
+|---|---|---|---|---|---|---|---|---|---|---|---|---|
+
+最早可得存档不一定是第一版。缺图/脚本/抓取失败不是页面当年没有对应功能的证据。Beta、正式发布与 PMF 不等价；PMF 如果只有推测就不用作确定分界点。截图用于结构与视觉，结构化文本用于内容/定价，两者互补。
+
+## 3. 早期用户证据表
+
+使用本包 [客户证据方法](early-customer-evidence.md)。必须把案例发布时间、早期身份、实际付费证据分别填清楚。案例没写付款就填未知，不为了满足“前十个”目标补猜客户身份。
+
+补一列“这是否改变了我方待访谈的人群/任务”，把发现转为招募假设。访谈和真实使用结果再修正假设。
+
+## 4. 博客与内容生产表
+
+| 观察窗口/样本方法 | 栏目/主题 | 样本数与占比 | 发布频次 | 内容结构/任务 | 作者与身份线索 | CTA/目标页 | 我方可复用做法 | 产能/费用估算假设 |
+|---|---|---|---|---|---|---|---|---|
+
+公开署名不能证明雇佣关系；按内容数量估算成本时写清单价、工作量和范围，不报告成对方实际营销费用。没有流量/转化数据时，只能判断内容布局，不能判断 ROI。
+
+## 5. 渠道与合作地图
+
+| 渠道 | 该渠道参考对象 | 历史/近期窗口 | 角色（官方/创始人/生态/创作者） | 内容类型/频次 | 合作形式及证据 | 可观测结果 | 数据缺口 | 预算区间及假设 | 我方适配度 |
+|---|---|---|---|---|---|---|---|---|---|
+
+- 历史研究与现在要执行的内容分列；近期可见帖子仍需核对规则，不能作为平台授权证据。
+- 平台观看量、工具估计流量、注册、激活、付款分别标记，不互相替代。
+- 没有赞助披露/官方确认时，写“提及”而非“付费合作”。
+- 对 KOL 的曝光/互动/报价加权只是筛选评分；真正 ROI 需要收入/利润与完整成本口径。
+
+Partner 表：
+
+| 官方页面 | 合作方企业 | 国家/市场 | Reseller/实施/集成/推荐/未知 | 服务人群 | 层级 | 公开商业联系入口 | 资格验证问题 | 下一步 |
+|---|---|---|---|---|---|---|---|---|
+
+列出线索不等于已联系或合作。外发邀约遵循当前用户授权，不能由研究任务自动触发。
+
+## 6. 价值主张与比较素材
+
+先写用户已有的替代方案和任务，再提取行业熟悉词汇，最后加入可证明的程度差异或能力差异。
+
+| 我方一句话 | 对象/任务 | 用户熟悉的词 | 比较性优势 | 事实证据及测试条件 | 目标用户如何复述 | 需要修改 |
+|---|---|---|---|---|---|---|
+
+对比测试使用可比较的条件和明确版本；缺证据不填叉。员工个人使用不写为企业采购或品牌授权。早期定价、免费额度、订阅权益与当前状态分开。
+
+## 7. 我方行动卡（本轮最重要交付）
+
+| 竞品证据 | 我方假设 | 可迁移条件 | 不照搬的部分 | 最小测试 | 预算/时间上限 | 负责人 | 指标与分母/观察窗 | 继续/修改/暂停条件 |
+|---|---|---|---|---|---|---|---|---|
+
+收尾回答：现在最值得先验证什么？哪些动作应暂缓？什么证据会推翻当前建议？
+
+
+## 反证与停止记录
+
+| 当前判断 | 对手真实优势 | 未知能力 | 会推翻判断的证据 | 调研时间上限 | 下一步验证 |
+|---|---|---|---|---|---|