project-retrospective

project-retrospective is a skill for Claude Code, Codex from TashanGKD/tashan-cursor-skills. It costs 108 tokens per session (7,206 once invoked), scanned A, original, MIT.

A project review workflow that examines the skills used during a completed project and records improvements for future work.

In plain words
What is it for?
Use it after project completion to review skill usage, capture gaps, and update the skill index.
Why use it?
It helps turn lessons from one project into updates to the team’s reusable skills and documentation.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Good fit Use it after project completion to review skill usage, capture gaps, and update the skill index.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/tashangkd/tashan-cursor-skills/project-retrospective
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add TashanGKD/tashan-cursor-skills --skill project-retrospective
Clone the repo
git clone --depth 1 https://github.com/TashanGKD/tashan-cursor-skills

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for project-retrospective

README.md
[![agentmods](https://agentmods.dev/badge/skills/tashangkd/tashan-cursor-skills/project-retrospective/github.svg)](https://agentmods.dev/skills/tashangkd/tashan-cursor-skills/project-retrospective)
Your own site
<a href="https://agentmods.dev/skills/tashangkd/tashan-cursor-skills/project-retrospective"><img src="https://agentmods.dev/badge/skills/tashangkd/tashan-cursor-skills/project-retrospective/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for project-retrospective

Your own site · 80×15
<a href="https://agentmods.dev/skills/tashangkd/tashan-cursor-skills/project-retrospective"><img src="https://agentmods.dev/badge/skills/tashangkd/tashan-cursor-skills/project-retrospective.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 108 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 7,206 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00108 $0.07206
Opus 5 $0.00054 $0.03603
Sonnet 5 $0.00022 $0.01441
Haiku 4.5 $0.00011 $0.00721

Measured 7d ago against content hash 4d522d8e2ac2, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

project-retrospective scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/project-retrospective/SKILL.md · 465 lines

How it starts

The opening of the file, as written. The whole thing — 465 lines — stays where its author put it; the contents beside it link to each section on GitHub.

项目级 Skill 复盘(project-retrospective)

单次经验用 skill-capture-closure(随时触发)。 项目结束后的系统性批量复盘用本 Skill(一次扫描全部)。 本 Skill 是外壳,skill-capture-closure 是内核——扫描由本 Skill 负责,每个具体更新动作仍走 skill-capture-closure 标准流程。


知识导航表(执行前必须理解的概念根)

本 Skill 的步骤设计基于以下文档的核心概念。若不理解这些概念,步骤无法正确执行。

层级 文档 需要理解的概念
D0 认知根(必读) _内部总控/认知结构/L1_系统性文档/系统架构思维维度/自进化智能体系统形式规范_v1.0.md 层1:B/K/S三类对象;层2:间隙(Gap)定义;层5:R/M/G基础设施 + M的批量处理逻辑;层7:D_op = {create, modify, delete, merge, split}
D3 规范参考 _内部总控/认知结构/L1_系统性文档/系统架构思维维度/Skill体系设计原则_v1.0.md §2.5 三型决策树(判断新建 Rule/Agent/Skill);§2.2/2.3/2.4 各类型本质
D4 运行时数据 .cursor/skills/skill-index/PENDING-EXPERIENCES.md + CO-BUILD-LOG.md + SKILL-INDEX.md G缓冲区(待处理Gap)+ 过程决策日志 + B-objects 注册表

核心概念速查: ① B-object(行为对象)= Skill/Rule/Agent,编码「在条件C下执行动作A」 ② K-object(知识对象)= 所有文档,编码「什么是真的」 ③ D_op = modify(修复已有)vs create(沉淀新建)—— 批量复盘需同时处理两类


激活后立即执行(顺序不可跳过)

Step 1  读取 Skill 总索引,了解所有现有 Skill
        Read: .cursor/skills/skill-index/SKILL-INDEX.md
        → 获取全量 Skill 列表,用于后续匹配

Step 2  获取「Skill 相关事件」清单(三路来源合并)
        
        【来源一:读取暂存文件(优先,已由 auto-experience-hook 自动积累)】
        Read: .cursor/skills/skill-index/PENDING-EXPERIENCES.md
        → 提取所有「🔲 待处理」条目
        → 捕捉:任务中的踩坑/新发现/步骤偏差(结果层)
        
        【来源二:读取协作建设日志(co-build-log 积累的决策轨迹)】
        Read: .cursor/skills/skill-index/CO-BUILD-LOG.md
        → 提取「待处理日志条目」区域的所有条目
        → 捕捉:任务中的推理链/决策轨迹/转折理由(过程层)
        → 重点关注类型 II(方案)和 III(转折)——这是提炼 Skill 的最高价值来源
        → 若文件不存在或无条目,跳过
        
        【来源三:补充扫描本次对话上下文(兜底,捕捉两个文件的遗漏)】
        回溯整个对话,识别以下信号(排除已在前两路中的条目,避免重复):
        
        【A. 调用了某个 Skill 且执行顺利】
           → 候选:状态升级为 ✅ 已验证(如果原来是 🔵 待验证)
        【B. 调用了某个 Skill 但执行有偏差/需要补充步骤】
           → 候选:更新该 Skill 的内容(修复/补充)
        【C. 遇到了某个场景,但没有合适的 Skill 覆盖,AI 凭感觉执行】
           → 候选:新建 Skill(Level 2)
        【D. 某个 Skill 的触发词不准(该触发没触发,或错误触发)】
           → 候选:更新 description 字段
        【E. 架构级发现(系统层面改进,不只是单个 Skill)】
           → 以下情况标为 E 类,不走 skill-updater,走 skill-designer 或规范修改流程:
           E1. AI 行为规律性不符预期,根因不在某个 Skill 的步骤里,需要新建 Rule
               → 候选:新建 alwaysApply Rule(.cursor/rules/)
           E1b. 已有 Rule 存在歧义/覆盖范围错误/需要修订
               → 候选:更新已有 Rule(走 skill-rule-修改规范)
           E2. 某类任务因上下文污染导致质量不稳定,子智能体隔离不够,需要新建
               → 候选:新建独立子智能体(.cursor/agents/)
           E2b. 已有 Agent 有设计缺陷/行为错误/描述不准确
               → 候选:更新已有 Agent(走 skill-rule-修改规范 Agent 分支)
           E3. 两个或多个 Skill/Agent/Rule 协作时有系统性摩擦或冲突
               → 候选:Level 3 集成型改造(走 skill-designer + 关卡A/B/C)
           E4. 需要重构多个组件的交互关系
               → 候选:Level 4 系统型调整(走 skill-designer 完整八步)
           E5. 发现某份规范/参考文档(非 .cursor/ 组件)内容有误、遗漏、需补充
               → 候选:更新规范文档(走 project-doc-versioning-guard 四步留痕)
               覆盖:开发规范手册 / 项目测试规范 / 技术架构文档 / 任何 _内部总控/开发规范/*.md 等
           E6. 【结构性设计问题】——AI 反复失败,但根因不是执行疏忽,而是规范/机制本身设计有缺陷
               判断标准(满足任一):
               ① 同类失败出现 ≥2 次,且每次复盘结论都是「AI 疏忽」
               ② 规范只在 Skill 里(被触发才生效),但该规范需要普遍约束
               ③ 重要文档/规范没有进入任何 D0 / alwaysApply Rule,AI 靠记忆才知道
               ④ 规范表述是「建议/应该」,实际需要「禁止/必须」级别的强制
               ⑤ 某个正确做法需要 AI 主动记住,而不是被机制强制
               → 候选:根据具体缺陷类型路由
               → 连接缺口 → 补 D0 / 新建 Rule(路径③)
               → 类型错误(Skill→Rule)→ 新建 alwaysApply Rule(路径③)
               → 规范质量 → 更新规范文档(路径⑦)
               → 执行力度 → 修改 Rule/Skill 的约束强度(路径⑥)
        
        → 将三路来源合并为一张完整清单,标注每条的类别(A/B/C/D/E)

Step 3  输出「待处理清单」,等用户确认
        格式:
        「本次项目系统复盘扫描结果:

          📋 Skill 更新类(A/B/C/D,走 skill-updater 路径):
          ① [Skill目录名] — [类型] — [一句话说明]
          ② ...

          🏗️ 架构级改进类(E,按子类走不同路径):
          ① [改进名称] — [E1/E1b/E2/E2b/E3/E4/E5] — [一句话说明]
          ② ...

          ⓪ 无需处理(执行符合预期):[列表]

          说明:
          - A/B/C/D 类:直接由 skill-updater 子智能体处理(快,无关卡)
          - E1/E2(新建Rule/Agent):走 skill-designer Level 2 流程(含关卡A+C)
          - E1b/E2b(更新已有Rule/Agent):走 skill-rule-修改规范(三问+备份+修改)
          - E3(多组件集成):走 skill-designer Level 3(含关卡A+B+C)
          - E4(系统重构):暂停,用户决策后开专题会话
          - E5(规范文档):走 project-doc-versioning-guard 四步留痕(快,无关卡)
          - E6(结构性设计问题):先诊断具体缺陷类型,再按连接缺口→③ / 类型错误→③ / 规范质量→⑦ / 执行力度→⑥ 分流

          请确认:[全部处理] [只处理Skill更新类] [只处理架构类] [跳过某项]」

        → 等用户确认后继续

Step 4  生成本次复盘的 ClosureCase 快照(13态状态机,显式追踪)
        
        创建文件:.cursor/skills/skill-index/closure-case-YYYYMMDD.yaml
        内容(状态机来自 user_global_rules_v1 闭环元模型):
        
        case_id: "project-retrospective-[日期]"
        goal: "本次复盘的目标"
        
        # 13态状态机(NEW→NORMALIZED→CONSTRAINED→PLANNED→RUNNING→
        #              WAITING_CHILDREN→MERGING→VERIFYING→QA_PENDING→
        #              DONE / BLOCKED / ESCALATED / ARCHIVED)
        status: "PLANNED"
        
        source_of_truth: ".cursor/skills/skill-index/SKILL-INDEX.md"
        current_stage: "implementation"   # context|model|plan|implementation|verification|delivery
        
        active_branches: [每个待处理 Skill 一个 branch_id]
        closed_branches: []
        repair_branches: []               # 验证失败后产生的 repair 分支
        artifacts: []
        
        acceptance_criteria:
          - "所有 🔲 待处理条目已处理"
          - "SKILL-INDEX 已同步更新"
          - "PENDING-EXPERIENCES 和 CO-BUILD-LOG 已清理"
          - "所有 repair_branches 已完成或已升级"
          - "QA 通过(无活动分支,无未解决冲突)"
        
        # 递归保护(来自 agent-io-contract.md)
        recursion_depth: 0
        branch_policy:
          max_children: 10              # 每次复盘最多10个并发 Skill 更新分支
          allowed: true
        escalation_policy:
          max_depth: 2
          escalate_on_block: true
        
        owner: "orchestrator"

Step 5  按条目类型路由分发处理(串行,A/B/C/D 与 E 类走不同路径)

        【路由规则(每条条目必须先判断类型再执行)】
        
        A/B/D 类(更新已有 Skill)→ 路径①:skill-updater 子智能体
        C 类(新建独立 Skill,Level 2)→ 路径②:skill-designer Level 2 流程
        E1 类(新建 Rule)→ 路径③:skill-designer Level 2 流程(创建 .cursor/rules/)
        E2 类(新建 Agent)→ 路径③:skill-designer Level 2 流程(创建 .cursor/agents/)
        E1b 类(更新已有 Rule)→ 路径⑥:skill-rule-修改规范(三问+备份+修改+变更记录)
        E2b 类(更新已有 Agent)→ 路径⑥:skill-rule-修改规范 Agent 分支(备份到 agents/history/)
        E3 类(多组件集成改造,Level 3)→ 路径④:skill-designer Level 3 流程(含关卡A+B+C)
        E4 类(系统重构,Level 4)→ 路径⑤:暂停,向用户确认规模后再执行
        E5 类(规范/参考文档更新)→ 路径⑦:project-doc-versioning-guard 四步(读版本→备份→修改→变更记录)
        E6 类(结构性设计问题)→ 先在 Step 5 内做二次分类,再按缺陷类型路由:
               · 连接缺口(缺 D0/Rule)→ 路径③(新建 Rule,skill-designer Level 2)
               · 类型错误(Skill→Rule)→ 路径③(新建 alwaysApply Rule)
               · 规范质量(表述歧义/覆盖错误)→ 路径⑦(更新规范文档)
               · 执行力度(建议→禁止)→ 路径⑥(修改已有 Rule/Skill 约束强度)
        
        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        【路径①:A/B/D 类 → skill-updater(串行,快速)】
        
        a. 明确宣告:「▶ 正在处理 [N/总数]:[Skill目录名] — [类型](路径①)」
        
        b. 构造统一任务信封(参见 skill-designer/agent-io-contract.md):
           task_id / case_id / branch_id / goal / scope /
           allowed_write_set / done_criteria / recursion_depth: 0
        
        c. 调用 /skill-updater,等待完成
        
        d. 解析输出:pass → 移入 closed_branches;fail → 创建 repair branch
        
        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        【路径②③:C/E1/E2 类 → skill-designer Level 2(含关卡A + 关卡C)】
        
        a. 明确宣告:「▶ 正在处理 [N/总数]:[组件名] — [类型](路径②/③,走 skill-designer)」
        
        b. 加载 skill-designer Skill,传入:
           - 需要创建的组件类型(Skill / Rule / Agent)
           - 来自复盘清单的需求描述
           - 当前 SKILL-INDEX 摘要
        
        c. skill-designer 走完 Step 3-7(双视角产品定义 → 编写 → 关卡A → 关卡C)
        
        d. 通过后由 skill-designer 执行 Step 8(部署)
        
        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        【路径④:E3 类 → skill-designer Level 3(含关卡A + B + C)】
        
        a. 明确宣告:「▶ 正在处理 [N/总数]:[改进名称] — E3(路径④,Level 3,含关卡B)」
        
        b. 加载 skill-designer Skill,Level 3 完整流程(Step 3-8)
        
        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        【路径⑥:E1b/E2b 类 → skill-rule-修改规范(更新已有 Rule 或 Agent)】

        a. 明确宣告:「▶ 正在处理 [N/总数]:[组件名] — [E1b/E2b](路径⑥,更新已有组件)」

        b. 执行 skill-rule-修改规范 完整流程:
           - 修改前三问(根因/影响/验证)
           - 备份:Rule → .cursor/rules/history/ / Agent → .cursor/agents/history/
           - 最小化修改
           - 追加变更记录

        c. 更新 SKILL-INDEX(若 Rule/Agent 在索引中有条目)

        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        【路径⑦:E5 类 → project-doc-versioning-guard(更新规范/参考文档)】

        a. 明确宣告:「▶ 正在处理 [N/总数]:[文档名] — E5(路径⑦,更新规范文档)」

        b. 执行 project-doc-versioning-guard 四步留痕:
           1. Read 文件头部 → 获取当前版本号
           2. cp 备份 → [所在目录]/history/[文件名]_v[版本]_YYYYMMDD.md
           3. 最小化修改(只改本次经验相关部分)
           4. 追加变更记录到文档末尾

        c. 本路径无需 skill-updater,主 Agent 直接执行

        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        【路径⑤:E4 类 → 暂停,用户决策】
        
        a. 输出:「⚠️ 发现 Level 4 系统型调整:[描述]
                  这将影响多个现有组件,建议开一个专门会话处理。
                  是否现在开始,还是加入 PENDING-SKILLS 后续规划?」
        
        b. 等待用户决策,不自行启动
        
        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
        所有条目处理完后,若有 repair branch,按原路径串行处理

Step 6  输出复盘报告,更新 ClosureCase,清理暂存区
        「📊 项目系统复盘完成

          Skill 更新(路径①):
          ✅ 更新已有 Skill:N 个(含版本升级)
          ✅ 状态升级:N 个(🔵→✅)
          🔧 repair 处理:N 个

          架构级改进(路径②③④⑥⑦):
          ✅ 新建 Skill(Level 2):N 个
          ✅ 新建 Rule:N 个
          ✅ 更新已有 Rule:N 个
          ✅ 新建 Agent:N 个
          ✅ 更新已有 Agent:N 个
          ✅ Level 3 集成改造:N 个
          ✅ 更新规范/参考文档:N 个

          待后续处理(路径⑤):
          📋 Level 4 系统调整:N 个(已加入 PENDING-SKILLS)

          ⏭ 跳过:N 个
          Skill 体系健康度:共 N 个,✅ X 个已验证,🔵 Y 个待验证」
        
        更新 ClosureCase 快照(closure-case-[日期].yaml):
        → status: "closed"
        → 所有处理完的 branch 移入 closed_branches
        
        清理暂存区(两个文件同步清理):
        → PENDING-EXPERIENCES.md:将所有 🔲 待处理 条目标记为 ✅ 已处理,移入归档区
        → CO-BUILD-LOG.md:将「待处理日志条目」中已提炼的条目移入「已处理归档」区

Read the full file on GitHub · 465 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 7d ago First seen · 465 lines · 108 tokens per session scan A 4d522d8e2ac2

Subscribe to this mod's changes

project-retrospective is a skill published in the GitHub repository TashanGKD/tashan-cursor-skills (20 stars, last pushed 5mo ago), licensed MIT. It adds 108 tokens to every session and 7,206 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other skills, from other repositories

product-adoption

Diagnose and design product adoption for onboarding, activation, behavior change, feature discovery, rollout, and sustained use after launch. Covers consumer, B2B, internal-tool, and public-service contexts with evidence-based decision points and cohort segmentation. Does not own acquisition campaigns or marketing…

magnus919/agent-skills · 114 tokens

team

Coordinate parallel work across multiple Cursor sessions via shared .omc/team/ state.

Jkudjo/oh-my-cursor · 17 tokens

template-helpers

Catalog of the reusable building blocks shipped with this Android template — BaseActivity/BaseFragment, the helpers/ extension files (navigation, lifecycle, toast, snackbar, dialogs, images, permissions, theme, locale, settings intents, date, delay), common/ (Firebase, network, observers), and…

orbitalsonic/AndroidPilot · 92 tokens

android-feature

Workflow for implementing a new Android feature, screen, fragment, activity, dialog, adapter, or enhancement in this XML/MVVM template. Use whenever the user asks to add, build, implement, or extend functionality — before writing any code. Covers planning, the MVVM/MVI skeleton, the Android configuration checklist…

orbitalsonic/AndroidPilot · 72 tokens

android-new-project

Workflow for turning a fresh copy of this Android template into a new app — renaming the package/namespace/applicationId, branding (icons, splash, palette, fonts), Firebase setup, signing config, stripping unused sample screens, and verifying the foundation. Use when the user says they are starting a new app…

orbitalsonic/AndroidPilot · 83 tokens

android-preflight

Final verification checklist to run before declaring Android work finished — build, both themes, string resources, lifecycle and leak risks, registered permissions and components, resource parity between values and values-night, and honest reporting of what was and was not verified. Use at the end of any feature, fix…

orbitalsonic/AndroidPilot · 79 tokens