[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M72W3AKSWPXPYEF1AP21E82A":3,"skill-same:01M72W3AKSWPXPYEF1AP21E82A":63},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":42,"commitSha":43,"skillDirectory":44,"forksCount":45,"repositoryHtmlUrl":46,"repositoryHomepageUrl":6,"ownerHtmlUrl":47,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":48,"install":49,"resources":62},"01M72W3AKSWPXPYEF1AP21E82A","reverification",null,"By-hand only — DO NOT auto-trigger. When the user types \u002Freverification (or asks\nfor \"reverify this\", \"independent verification\", \"LLM-uncheatable check\",\n\"第三方再验证\", \"用 repo 之外的方法验证\"), re-verify recently-completed work\nusing verification methods that **existed before this repo existed** and that an\nLLM agent cannot fabricate. Pick 3-5 verifications from a 10-family menu\n(hardware \u002F cryptographic \u002F blockchain \u002F external real-system \u002F mathematical \u002F\ntemporal \u002F content authenticity \u002F physical \u002F network \u002F third-party trust),\ndispatch them via real tools that capture raw evidence, and report PASS \u002F FAIL \u002F\nINCONCLUSIVE based ONLY on captured evidence — never on internal reasoning.\nRefuse to fall back on TeamBrain's bespoke verifiers (judge harness, fastprobe,\n\u002Freview, feature-verification gates) because those did NOT exist before the repo.","libz-renlab-ai\u002FMatrix-Viki","libz-renlab-ai","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F277592717?v=4","01M71ENHYC54KKYC8Q3KDYG8S2","01M72W199DPMZF71MDAXQMVEY2",1,"MIT","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Freverification\u002FSKILL.md",[],[],"2026-11-10 22:39:42",false,"---\nname: reverification\ndescription: |\n  By-hand only — DO NOT auto-trigger. When the user types \u002Freverification (or asks\n  for \"reverify this\", \"independent verification\", \"LLM-uncheatable check\",\n  \"第三方再验证\", \"用 repo 之外的方法验证\"), re-verify recently-completed work\n  using verification methods that **existed before this repo existed** and that an\n  LLM agent cannot fabricate. Pick 3-5 verifications from a 10-family menu\n  (hardware \u002F cryptographic \u002F blockchain \u002F external real-system \u002F mathematical \u002F\n  temporal \u002F content authenticity \u002F physical \u002F network \u002F third-party trust),\n  dispatch them via real tools that capture raw evidence, and report PASS \u002F FAIL \u002F\n  INCONCLUSIVE based ONLY on captured evidence — never on internal reasoning.\n  Refuse to fall back on TeamBrain's bespoke verifiers (judge harness, fastprobe,\n  \u002Freview, feature-verification gates) because those did NOT exist before the repo.\nallowed-tools:\n  - Bash\n  - Read\n  - WebFetch\n  - WebSearch\n  - Grep\n  - Glob\n  - Agent\ntriggers:\n  - \u002Freverification\n  - reverify this\n  - reverify the work\n  - independent verification\n  - LLM-uncheatable check\n  - pre-existing verification\n  - 第三方再验证\n  - 用 repo 之外的方法验证\n  - 不让 LLM 作弊的验证\n---\n\n# \u002Freverification — 第三方再验证（pre-existing & LLM-uncheatable）\n\n```\n   __                                                          .---.\n _(o o)\u003C   呷呷~ 鸭鸭是 by-hand only 的再验证向导            ( raw  )\n  \\   )     不写死步骤、不调 repo 自家工具                   ( evi  )\n   `-'      由 LLM 现场从「repo 还没出生时就存在」的          ( dence )\n            验证宇宙里挑 3-5 个，跑出 raw evidence，          `-----'\n            再用 evidence 反推 PASS \u002F FAIL                       ↑\n                                                                 │\n                                                          外部物理 \u002F 密码学\n                                                          \u002F 网络 \u002F 第三方\n```\n\n## 灵魂条款（不要违反）\n\n这只 skill 不写死任何脚本、任何固定命令、任何固定顺序。**调用时**的 LLM 自己负责挑、自己负责跑、自己负责诚实承认抓不到证据时的失败。\n\n为什么这样？因为 TeamBrain 自家的验证体系（judge harness \u002F fastprobe \u002F `\u002Freview` \u002F feature-verification gates）都是 repo 出生**之后**才有的，LLM 完全可以理解它们的内部逻辑并伪造看起来合理的 output。「pre-existing & uncheatable」的本质是：**验证结果依赖 LLM 触碰不到的外部现实** —— 它不知道某个信息，而这个信息必须通过物理世界、密码学机制、外部网络或第三方信任才能获得。\n\n优先级链：**硬件锚定 > 密码学承诺 > 外部真实系统 > 数学决定性证明**。\n\n## 验证宇宙（10 家族菜单 —— LLM 现场选，不是清单）\n\n| # | 家族 | 代表手段 | 为什么 LLM 作弊不了 |\n|---|------|---------|---------------------|\n| 1 | Cryptographic \u002F hash \u002F signature | Merkle 证明、RSA\u002FECDSA、PKI 链、RFC 3161 时间戳、HMAC、ZKP、homomorphic hash | 私钥 \u002F 共享密钥 \u002F 真实数据 hash 链 LLM 拿不到 |\n| 2 | Hardware-backed security | TPM EK、SGX\u002FTrustZone enclave、HSM、Secure Boot、TEE remote attestation | 芯片级信任根，飞地内存不可读 |\n| 3 | Blockchain \u002F distributed consensus | PoW 链历史、智能合约状态、ENS、IPFS CID、Sigstore\u002FRekor 签名 commit | 共识 \u002F 透明日志需多方算力，LLM 单方面改不动 |\n| 4 | External real-system APIs | 天气、ADS-B 航班、交易所 order book、卫星图像、IoT 传感器、医疗 telemetry、SCADA | 真实物理信号 \u002F 实时下行数据，LLM 接不到 |\n| 5 | Mathematical decidable proofs | SAT\u002FSMT solver、Coq\u002FLean、DH\u002FECDH、KZG commitment、SNARK | 机器证明 + kernel 校验，LLM 绕不过 |\n| 6 | Temporal \u002F ordering | NTP、Lamport 向量、FIFO 队列、VDF（verifiable delay function）、可验证延迟编码 | 真实时间流逝，LLM 在推理时压缩不了 |\n| 7 | Content \u002F media authenticity | EXIF、相机 RAW、视频关键帧 hash、音频声纹、FASTQ DNA reads、质谱原始数据 | 仪器硬件签名 \u002F 物理传感器原始噪声分布 |\n| 8 | Physical \u002F geographic | GPS+原子钟、RFID\u002FNFC PUF 标签、邮政追踪、SWIFT 电文、公证档案、法院案号 | 卫星定位 \u002F 物理不可克隆函数 \u002F 政府登记 |\n| 9 | Network \u002F transport | TCP seq number、TLS handshake 机密、DNSSEC、mTLS 双向证书 | 三次握手随机数 \u002F pre-master secret LLM 预测不到 |\n| 10 | Third-party trust anchors | Sigstore\u002FRekor 透明日志、CWT、OCSP 吊销、W3C VC、Slack\u002FDiscord server 时间戳 | 多个独立信任源同时伪造的成本 |\n\n完整 55+ 条目录参见用户 `claudefast -p \"list ... LLM cheat ...\"` 的输出（每次调用重新生成，不要在这里硬列）。\n\n## 调用流程（不是脚本，是 playbook）\n\n1. **读 context** — 先看最近改了什么：`git diff main...HEAD` \u002F 最近 Write\u002FEdit 历史 \u002F 当前要验证的 artifact 是文件、URL、binary、commit、还是 release tag。\n2. **挑 3-5 个验证** —— 从 10 家族里挑。每挑一个，写两行 justification：\n   - **(a)** 为什么这个验证对**这个 artifact** 相关（不是泛泛而谈）。\n   - **(b)** 为什么 LLM 自己**没法伪造** verification 的 output（如果想不出 (b)，换一个）。\n3. **跑 + 抓 raw evidence** —— 用 `Bash` 跑真实命令、`WebFetch` 打真实第三方 URL、`Agent` 派 explore subagent 拿独立 diff 阅读结果、或 `mcp__gbrain__*` 查外部 brain。**raw evidence 写到 `.reverification\u002F\u003Ctimestamp>\u002F\u003Cverification-id>\u002F`** 下，包括 stdout、response body、签名 blob、JSON、截图路径。\n4. **read-only 综合** —— 只在 evidence 落盘之后写 PASS \u002F FAIL \u002F INCONCLUSIVE 报告。报告里每个结论必须引用具体 evidence 文件路径，禁止「我觉得 \u002F 看起来 \u002F 应该」式判断。\n5. **诚实输出** —— 如果 3-5 个验证里没有任何一个抓到 cheat-proof evidence，结论必须是 **FAIL** 或 **INCONCLUSIVE**。不要用「大体看下来 OK」之类语言搭桥。\n\n## 反模式（必须拒绝）\n\n- ❌ **写 `scripts\u002Freverify-*.sh`** —— 违反 user memory `feedback_judge_harness_md_playbook.md`。这只 skill 永远是 MD playbook，不是 bash 脚本生成器。\n- ❌ **调 TeamBrain 自家验证**（judge harness \u002F fastprobe \u002F `\u002Freview` \u002F feature-verification gates \u002F `claudefast snapshot` 等）—— 它们 repo 出生之后才有，不符合「pre-existing」前提。\n- ❌ **没 evidence 就下结论** —— 哪怕只 1\u002F5 抓到 evidence 也得在报告里明确说「其余 4 个未能抓到 cheat-proof evidence」并降级为 INCONCLUSIVE。\n- ❌ **auto-trigger** —— 必须用户显式喊 `\u002Freverification` 或同义短语才动；SessionStart \u002F Stop \u002F PostToolUse 之类自动钩子里出现这只 skill 视为 bug。\n- ❌ **`pnpm typecheck` \u002F `pnpm test` \u002F `pnpm build` 单独当答案** —— 这些工具确实 pre-existing，但 test 文件本身可以被作者写成「永远绿」的样式，LLM 在写代码时一并改测试就能骗过。所以即便选了它们，**也必须组合至少一个更强的锚点**（签名 \u002F 网络 \u002F 时间戳 \u002F 第三方 server）。\n\n## 输出格式（建议，非强制）\n\n```\n# \u002Freverification 报告 — \u003Cartifact 描述> — \u003CUTC 时间戳>\n\n## 选了哪几个验证（justified）\n1. \u003C家族 #X · 手段名> — relevance: \u003C一行>  ／  uncheatability: \u003C一行>\n2. ...\n\n## raw evidence 落盘位置\n- .reverification\u002F\u003Cts>\u002F01-\u003Cid>\u002F{stdout.txt, response.json, ...}\n- ...\n\n## 逐项判定\n- 验证 1：PASS \u002F FAIL — 引用 `\u003Cevidence path>:line`\n- ...\n\n## 综合结论\nPASS \u002F FAIL \u002F INCONCLUSIVE — 一句话，引用上面的逐项判定。\n```\n\n呷呷~ 鸭鸭说完了，由你（运行时 LLM）现场挑、现场抓证据、现场承担诚实义务。(>ω\u003C)\n","# \u002Freverification — 第三方再验证（pre-existing & LLM-uncheatable）\n\n```\n   __                                                          .---.\n _(o o)\u003C   呷呷~ 鸭鸭是 by-hand only 的再验证向导            ( raw  )\n  \\   )     不写死步骤、不调 repo 自家工具                   ( evi  )\n   `-'      由 LLM 现场从「repo 还没出生时就存在」的          ( dence )\n            验证宇宙里挑 3-5 个，跑出 raw evidence，          `-----'\n            再用 evidence 反推 PASS \u002F FAIL                       ↑\n                                                                 │\n                                                          外部物理 \u002F 密码学\n                                                          \u002F 网络 \u002F 第三方\n```\n\n## 灵魂条款（不要违反）\n\n这只 skill 不写死任何脚本、任何固定命令、任何固定顺序。**调用时**的 LLM 自己负责挑、自己负责跑、自己负责诚实承认抓不到证据时的失败。\n\n为什么这样？因为 TeamBrain 自家的验证体系（judge harness \u002F fastprobe \u002F `\u002Freview` \u002F feature-verification gates）都是 repo 出生**之后**才有的，LLM 完全可以理解它们的内部逻辑并伪造看起来合理的 output。「pre-existing & uncheatable」的本质是：**验证结果依赖 LLM 触碰不到的外部现实** —— 它不知道某个信息，而这个信息必须通过物理世界、密码学机制、外部网络或第三方信任才能获得。\n\n优先级链：**硬件锚定 > 密码学承诺 > 外部真实系统 > 数学决定性证明**。\n\n## 验证宇宙（10 家族菜单 —— LLM 现场选，不是清单）\n\n| # | 家族 | 代表手段 | 为什么 LLM 作弊不了 |\n|---|------|---------|---------------------|\n| 1 | Cryptographic \u002F hash \u002F signature | Merkle 证明、RSA\u002FECDSA、PKI 链、RFC 3161 时间戳、HMAC、ZKP、homomorphic hash | 私钥 \u002F 共享密钥 \u002F 真实数据 hash 链 LLM 拿不到 |\n| 2 | Hardware-backed security | TPM EK、SGX\u002FTrustZone enclave、HSM、Secure Boot、TEE remote attestation | 芯片级信任根，飞地内存不可读 |\n| 3 | Blockchain \u002F distributed consensus | PoW 链历史、智能合约状态、ENS、IPFS CID、Sigstore\u002FRekor 签名 commit | 共识 \u002F 透明日志需多方算力，LLM 单方面改不动 |\n| 4 | External real-system APIs | 天气、ADS-B 航班、交易所 order book、卫星图像、IoT 传感器、医疗 telemetry、SCADA | 真实物理信号 \u002F 实时下行数据，LLM 接不到 |\n| 5 | Mathematical decidable proofs | SAT\u002FSMT solver、Coq\u002FLean、DH\u002FECDH、KZG commitment、SNARK | 机器证明 + kernel 校验，LLM 绕不过 |\n| 6 | Temporal \u002F ordering | NTP、Lamport 向量、FIFO 队列、VDF（verifiable delay function）、可验证延迟编码 | 真实时间流逝，LLM 在推理时压缩不了 |\n| 7 | Content \u002F media authenticity | EXIF、相机 RAW、视频关键帧 hash、音频声纹、FASTQ DNA reads、质谱原始数据 | 仪器硬件签名 \u002F 物理传感器原始噪声分布 |\n| 8 | Physical \u002F geographic | GPS+原子钟、RFID\u002FNFC PUF 标签、邮政追踪、SWIFT 电文、公证档案、法院案号 | 卫星定位 \u002F 物理不可克隆函数 \u002F 政府登记 |\n| 9 | Network \u002F transport | TCP seq number、TLS handshake 机密、DNSSEC、mTLS 双向证书 | 三次握手随机数 \u002F pre-master secret LLM 预测不到 |\n| 10 | Third-party trust anchors | Sigstore\u002FRekor 透明日志、CWT、OCSP 吊销、W3C VC、Slack\u002FDiscord server 时间戳 | 多个独立信任源同时伪造的成本 |\n\n完整 55+ 条目录参见用户 `claudefast -p \"list ... LLM cheat ...\"` 的输出（每次调用重新生成，不要在这里硬列）。\n\n## 调用流程（不是脚本，是 playbook）\n\n1. **读 context** — 先看最近改了什么：`git diff main...HEAD` \u002F 最近 Write\u002FEdit 历史 \u002F 当前要验证的 artifact 是文件、URL、binary、commit、还是 release tag。\n2. **挑 3-5 个验证** —— 从 10 家族里挑。每挑一个，写两行 justification：\n   - **(a)** 为什么这个验证对**这个 artifact** 相关（不是泛泛而谈）。\n   - **(b)** 为什么 LLM 自己**没法伪造** verification 的 output（如果想不出 (b)，换一个）。\n3. **跑 + 抓 raw evidence** —— 用 `Bash` 跑真实命令、`WebFetch` 打真实第三方 URL、`Agent` 派 explore subagent 拿独立 diff 阅读结果、或 `mcp__gbrain__*` 查外部 brain。**raw evidence 写到 `.reverification\u002F\u003Ctimestamp>\u002F\u003Cverification-id>\u002F`** 下，包括 stdout、response body、签名 blob、JSON、截图路径。\n4. **read-only 综合** —— 只在 evidence 落盘之后写 PASS \u002F FAIL \u002F INCONCLUSIVE 报告。报告里每个结论必须引用具体 evidence 文件路径，禁止「我觉得 \u002F 看起来 \u002F 应该」式判断。\n5. **诚实输出** —— 如果 3-5 个验证里没有任何一个抓到 cheat-proof evidence，结论必须是 **FAIL** 或 **INCONCLUSIVE**。不要用「大体看下来 OK」之类语言搭桥。\n\n## 反模式（必须拒绝）\n\n- ❌ **写 `scripts\u002Freverify-*.sh`** —— 违反 user memory `feedback_judge_harness_md_playbook.md`。这只 skill 永远是 MD playbook，不是 bash 脚本生成器。\n- ❌ **调 TeamBrain 自家验证**（judge harness \u002F fastprobe \u002F `\u002Freview` \u002F feature-verification gates \u002F `claudefast snapshot` 等）—— 它们 repo 出生之后才有，不符合「pre-existing」前提。\n- ❌ **没 evidence 就下结论** —— 哪怕只 1\u002F5 抓到 evidence 也得在报告里明确说「其余 4 个未能抓到 cheat-proof evidence」并降级为 INCONCLUSIVE。\n- ❌ **auto-trigger** —— 必须用户显式喊 `\u002Freverification` 或同义短语才动；SessionStart \u002F Stop \u002F PostToolUse 之类自动钩子里出现这只 skill 视为 bug。\n- ❌ **`pnpm typecheck` \u002F `pnpm test` \u002F `pnpm build` 单独当答案** —— 这些工具确实 pre-existing，但 test 文件本身可以被作者写成「永远绿」的样式，LLM 在写代码时一并改测试就能骗过。所以即便选了它们，**也必须组合至少一个更强的锚点**（签名 \u002F 网络 \u002F 时间戳 \u002F 第三方 server）。\n\n## 输出格式（建议，非强制）\n\n```\n# \u002Freverification 报告 — \u003Cartifact 描述> — \u003CUTC 时间戳>\n\n## 选了哪几个验证（justified）\n1. \u003C家族 #X · 手段名> — relevance: \u003C一行>  ／  uncheatability: \u003C一行>\n2. ...\n\n## raw evidence 落盘位置\n- .reverification\u002F\u003Cts>\u002F01-\u003Cid>\u002F{stdout.txt, response.json, ...}\n- ...\n\n## 逐项判定\n- 验证 1：PASS \u002F FAIL — 引用 `\u003Cevidence path>:line`\n- ...\n\n## 综合结论\nPASS \u002F FAIL \u002F INCONCLUSIVE — 一句话，引用上面的逐项判定。\n```\n\n呷呷~ 鸭鸭说完了，由你（运行时 LLM）现场挑、现场抓证据、现场承担诚实义务。(>ω\u003C)\n",{"name":5,"triggers":23,"description":33,"allowed-tools":34},[24,25,26,27,28,29,30,31,32],"\u002Freverification","reverify this","reverify the work","independent verification","LLM-uncheatable check","pre-existing verification","第三方再验证","用 repo 之外的方法验证","不让 LLM 作弊的验证","By-hand only — DO NOT auto-trigger. When the user types \u002Freverification (or asks\nfor \"reverify this\", \"independent verification\", \"LLM-uncheatable check\",\n\"第三方再验证\", \"用 repo 之外的方法验证\"), re-verify recently-completed work\nusing verification methods that **existed before this repo existed** and that an\nLLM agent cannot fabricate. Pick 3-5 verifications from a 10-family menu\n(hardware \u002F cryptographic \u002F blockchain \u002F external real-system \u002F mathematical \u002F\ntemporal \u002F content authenticity \u002F physical \u002F network \u002F third-party trust),\ndispatch them via real tools that capture raw evidence, and report PASS \u002F FAIL \u002F\nINCONCLUSIVE based ONLY on captured evidence — never on internal reasoning.\nRefuse to fall back on TeamBrain's bespoke verifiers (judge harness, fastprobe,\n\u002Freview, feature-verification gates) because those did NOT exist before the repo.\n",[35,36,37,38,39,40,41],"Bash","Read","WebFetch","WebSearch","Grep","Glob","Agent","main","2ff5f3d80eff5adb8e7166cfb1182707e8098af2",".claude\u002Fskills\u002Freverification",0,"https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai","2026-05-20 13:23:26",{"source":8,"skillName":5,"commands":50,"prompt":60,"zipAvailable":61},[51,54,57],{"tool":52,"command":53},"npx","npx skills add libz-renlab-ai\u002FMatrix-Viki --skill reverification",{"tool":55,"command":56},"bunx","bunx skills add libz-renlab-ai\u002FMatrix-Viki --skill reverification",{"tool":58,"command":59},"pnpm","pnpm dlx skills add libz-renlab-ai\u002FMatrix-Viki --skill reverification","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Freverification\u002FSKILL.md",true,[],[64,71,78,85,92,97],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":18},"01M72W66NEM357BYZZ2KQWKMK5","zoom-out","Tell the agent to zoom out and give broader context or a higher-level perspective. Use when you're unfamiliar with a section of code or need to understand how it fits into the bigger picture.","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fzoom-out\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":18},"01M72W64QAPK9T1R0PQAA5FVGP","triage","Triage issues through a state machine driven by triage roles. Use when user wants to create an issue, triage issues, review incoming bugs or feature requests, prepare issues for an AFK agent, or manage issue workflow.","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Ftriage\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":18},"01M72W62GXS5BKZ9HDKWGT8R9C","to-prd","Turn the current conversation context into a PRD and publish it to the project issue tracker. Use when user wants to create a PRD from the current context.","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fto-prd\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":18},"01M72W60EP8VC16C50PXYXJC5D","to-issues","Break a plan, spec, or PRD into independently-grabbable issues on the project issue tracker using tracer-bullet vertical slices. Use when user wants to convert a plan into issues, create implementation tickets, or break down work into issues.","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fto-issues\u002FSKILL.md",[],[],{"skillNo":93,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":18},"01M72W5V80MPEMH126EPFRMYM5","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Freverification\u002FSKILL.md",[],[],{"skillNo":98,"name":99,"nameZh":6,"description":100,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":101,"categories":102,"keywords":103,"publishedAt":18},"01M72W5RB0B7C5YWPMYQFQBTFP","repo-issues-status","TeamBrain 仓库**全量 issue 全景**报告（all-scope only）。仅当用户问\"status of all issues\" \u002F \"show me all issues\" \u002F \"what issues are open\" \u002F \"all open issues\" \u002F \"issue 状态（全景）\" \u002F \"所有 issue 状态\" \u002F \"issue 全景\" \u002F \"仓库 issue 概览\" \u002F \"还有哪些 open issue\" 这类**仓库级、多 issue、概览类**问题时触发。**不要在以下情况触发**：(a) 用户问单个 issue 的状态（\"what's the status of issue","https:\u002F\u002Fgithub.com\u002Flibz-renlab-ai\u002FMatrix-Viki\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Frepo-issues-status\u002FSKILL.md",[],[]]