fancy-claude-code

Execute Story design review for an Epic or single Story using parallel three-layer adversarial analysis (Structure Hunter, Consistency Checker, Contract Auditor) with four-bucket triage, and save review summary to a structured result file. Use when user mentions 'SR', 'story review', 'story design review', 'epic review', 'story check', 'design review', 'Story 审查', 'SR 审查', '设计审查', 'Story 设计审查', 'Epic 审查', '审查 Story', '审查 Epic', or wants to review Story design documents before development. Capable of dual-granularity review (Epic-level or single-Story), parallel three-layer adversarial analysis, four-bucket triage, auto-detecting review history, batch processing, and generating structured review result documents with round numbering.

Evaluate Story design review results and generate a structured evaluation document. Use when user mentions 'SR evaluate', 'SR evaluation', 'evaluate SR', 'story review evaluation', 'design review evaluation', '评估 SR', '评估审查结果', 'SR 评估', '设计审查评估', 'Story 审查评估', '评估 Story 审查', or wants to assess Story design review findings. Capable of reading latest review results, assessing finding validity with source and bucket awareness, and producing evaluation documents with round numbering.

Execute Story document revisions based on SR evaluation conclusions and append revision summary to the evaluation document. Use when user mentions 'SR fix', 'SR revise', 'story revision', 'apply SR fixes', 'story review fix', 'design fix', '执行修订', 'SR 修订', '修订 Story', '执行 SR 修正', 'Story 设计修订', '修订 Story 文档', or wants to implement revisions from SR evaluation. Capable of reading evaluation conclusions, executing targeted document revisions, and recording revision summaries in evaluation documents.

Execute cross-LLM code review for a Story using parallel adversarial review layers (Blind Hunter, Edge Case Hunter, Acceptance Auditor) with structured triage, and save review summary to a structured result file. Use when user mentions 'CR', 'code review', 'crossllm review', 're-review', 'code review summary', '代码审查', 'CR 审查', '跨模型审查', '代码复审', '代码评审', '审查代码', or wants to review Story code changes. Capable of first-round and subsequent-round code reviews, parallel three-layer adversarial analysis, four-bucket triage, auto-detecting review history, and generating structured review result documents with round numbering.

Evaluate code review results from a cross-LLM review round and generate a structured evaluation document. Use when user mentions 'CR evaluate', 'CR evaluation', 'evaluate CR', 'review assessment', 'code review evaluation', '评估审查结果', '评估 CR', '评估 CR 结果', 'CR 评估', '审查结果评估', 'CR 结果评估', '代码审查评估', or wants to assess code review findings. Capable of reading latest review results, assessing finding validity, and producing evaluation documents with round numbering.

Execute code fixes based on code review evaluation conclusions and append fix summary to the evaluation document. Use when user mentions 'CR fix', 'CR repair', 'execute fix', 'apply CR fixes', 'code review fix', '执行修复', 'CR 修复', '修复 CR 问题', '执行 CR 修正', '代码审查修复', or wants to implement fixes from CR evaluation. Capable of reading evaluation conclusions, executing targeted code fixes, and recording fix summaries in evaluation documents.

Analyze historical code review, evaluation, and fix records to extract reusable development guidelines and best practices for project-level documentation. Use when user mentions 'extract CR rules', 'CR summary', 'extract CR guidelines', 'CR best practices', 'code review lessons', '提炼 CR 规则', 'CR 总结', '提取 CR 最佳实践', 'CR 经验总结', '代码审查规则提炼', '总结 CR 经验', or wants to distill patterns from CR history. Capable of analyzing multi-round CR findings, identifying recurring issues, and proposing updates to global project documents like project-context.md and architect.md.

Manage CR TODO backlog — add, check, resolve, and list deferred improvement items from code reviews. Use when user mentions 'CR TODO', 'add TODO', 'check TODO', 'resolve TODO', 'list TODO', 'CR backlog', 'extract todos', 'CR 待办', '添加 TODO', '检查 TODO', '解决 TODO', '查看 TODO', 'CR 延迟事项', '批量提取 TODO', or wants to track non-blocking improvements identified during code reviews.

Update Story status to Done and sync workflow tracking documents after CR approval. Use when user mentions 'CR done', 'CR approved', 'CR complete', 'story done', 'mark done', 'close story', 'CR 完成', 'CR 通过', '标记完成', '关闭 Story', 'Story 完成', '更新状态为 Done', 'CR 收尾', or wants to finalize a Story after code review approval. Capable of updating Story status, sprint-status.yaml, and bmm-workflow-status.yaml in one atomic operation.

Audit code changes against acceptance criteria (AC) from Story specifications, reporting violations, deviations, and unimplemented behaviors as a structured Markdown list. Use when user mentions 'acceptance audit', 'AC audit', 'AC review', 'acceptance criteria check', 'spec compliance', 'acceptance auditor', '验收审计', 'AC 审查', '验收标准检查', '验收对照', '规格合规审查', or needs to verify code against Story acceptance criteria. Capable of cross-referencing AC items with code changes, identifying spec violations and gaps, outputting structured findings with AC references and code evidence.

Perform an adversarial review of code diffs, specs, stories, or any artifact, finding at least ten issues and outputting a Markdown findings list. Use when user mentions 'adversarial review', 'cynical review', 'critical review', 'blind review', 'blind hunter', 'review content', '对抗式审查', '批判性审查', '盲审', '找问题', '挑毛病', '审查内容', or asks for a harsh/skeptical review of any content. Capable of reviewing diffs, specs, story files, documents, and arbitrary artifacts with extreme skepticism, outputting structured findings.

Exhaustively enumerate every branching path and boundary condition in code diffs, files, or functions, reporting only unhandled edge cases as a JSON array. Use when user mentions 'edge case', 'edge case hunter', 'boundary conditions', 'path analysis', 'unhandled cases', 'exhaustive review', '边界条件', '边缘情况', '路径分析', '未处理情况', '边界审查', '穷举分析', or needs orthogonal mechanical path-tracing review. Capable of scanning diffs, full files, and functions; outputs structured JSON with location, trigger condition, guard snippet, and potential consequence for each finding.

Summarize recent project changes, generate standardized Git commit messages following Conventional Commits specification, commit to local repository, and optionally push to remote. Use when user mentions 'git commit', 'commit', 'gc', 'gcp', 'commit and push', 'push', 'conventional commit', 'submit code', 'commit message', '提交代码', '提交到 Git', 'Git 提交', '生成提交信息', '规范提交', '约定式提交', '帮我提交', '代码提交', '提交并推送', '推送到远程', '同步远程', or wants to commit staged/unstaged changes with well-formatted messages. Capable of multi-file change summarization, Conventional Commits type classification, scope extraction, breaking change detection, automated git commit execution, and conditional remote push with safety checks.

Interactive Skill development assistant that creates complete Agent Skills packages through structured dialogue. Use when user mentions 'create skill', 'new skill', 'build skill', 'generate skill', 'skill creator', 'skill development', 'skill pack', '创建 skill', '新建技能', '生成技能', '开发技能', '技能创建', '创建技能包', '帮我做个 Skill', or wants to create SKILL.md files, agent skills, or automate workflow packaging. Capable of progressive disclosure architecture design, five workflow pattern matching, YAML frontmatter generation, reference document structuring, Python/Shell script scaffolding, and trigger testing guidance.

Validate Agent Skills against Anthropic open standard specifications, checking YAML frontmatter, file naming, description quality, version consistency, and content constraints. Use when user mentions 'skills-lint', 'lint skill', 'validate skill', 'check skill', 'skill lint', 'skill check', 'skill validation', '检查 Skill', '检查 Skill 规范', '验证技能', 'Skill 规范检查', 'Skill 合规检查', '技能验证', '检查技能格式', or wants to audit an existing Skill for compliance. Capable of YAML violation detection, naming convention verification, bilingual trigger keyword coverage analysis, version mismatch identification, forbidden file scanning, and structured report generation.

Manage Skill version iteration by analyzing changes, determining semantic version bumps, updating metadata.version in SKILL.md and SKILL.en.md, and maintaining CHANGELOG.md records. Use when user mentions 'skills-upgrade', 'upgrade skill', 'bump version', 'update version', 'skill upgrade', 'version bump', 'new version', 'release version', '升级版本', '更新版本号', '版本升级', 'Skill 升级', '发布新版本', '更新 CHANGELOG', '升级 Skill 版本', or wants to record changes after modifying a Skill. Capable of semantic versioning analysis (MAJOR/MINOR/PATCH), CHANGELOG entry generation with categorized changes, multi-copy synchronization between forge/ and installed skill roots, and post-upgrade lint recommendation.

编写和维护架构决策记录（ADR），遵循技术决策文档化的最佳实践。适用于记录重大技术决策、审查过往架构选择或建立决策流程。

实现经过验证的后端架构模式，包括 Clean Architecture、Hexagonal Architecture 和 Domain-Driven Design。适用于设计复杂后端系统或重构现有应用以提升可维护性。

Guide GitLab fork repository workflows with upstream setup, fork sync via rebase, conventional commits, and merge request submission. Use when user mentions sync fork, rebase upstream, merge request, MR, upstream, fork workflow, 派生仓库, 合并回原仓库, 同步上游, or needs to contribute from a fork back to the original repository. Do NOT use for general Git branching, regular merge operations, or non-fork repository workflows. Capable of safety checks, feature branch creation, conflict-resolution guidance, force-push handling after rebase, and GitLab MR preparation.

精通 JPA/Hibernate - 实体设计、查询、事务、性能优化

构建生产级 Spring Boot 应用 - REST API、Security、Data、Actuator

测试 Java 应用 - JUnit 5、Mockito、集成测试、TDD 模式

Spring Boot 架构模式、REST API 设计、分层服务、数据访问、缓存、异步处理和日志。适用于 Java Spring Boot 后端开发。