MCP Server Builder
anthropics/skills
Guides the design and implementation of Model Context Protocol servers in TypeScript or Python, from tool naming and error messages to evaluation.
M-1.5 review skill — 在 patch 跟 contract 之间找 finding，produce Finding 一等对象。
Install Review
$ npx skills add Towow-ai/Flowness --skill review -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install Towow-ai/Flowness review --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/Towow-ai/Flowness.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/review .claude/skills/review && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "review" agent skill from https://github.com/Towow-ai/Flowness/tree/main/.claude/skills/review into .claude/skills/review/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "review", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/Towow-ai/Flowness/tree/main/.claude/skills/reviewType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add Towow-ai/Flowness --skill review -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install Towow-ai/Flowness review --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Towow-ai/Flowness.git skills-src && mkdir -p .agents/skills && cp -r skills-src/.claude/skills/review .agents/skills/review && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "review" agent skill from https://github.com/Towow-ai/Flowness/tree/main/.claude/skills/review into .agents/skills/review/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "review", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add Towow-ai/Flowness --skill review -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install Towow-ai/Flowness review --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Towow-ai/Flowness.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/.claude/skills/review .cursor/skills/review && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "review" agent skill from https://github.com/Towow-ai/Flowness/tree/main/.claude/skills/review into .cursor/skills/review/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "review", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/Towow-ai/Flowness.git --path .claude/skills/review--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add Towow-ai/Flowness --skill review -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install Towow-ai/Flowness review --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Towow-ai/Flowness.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/.claude/skills/review .gemini/skills/review && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "review" agent skill from https://github.com/Towow-ai/Flowness/tree/main/.claude/skills/review into .gemini/skills/review/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "review", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install Towow-ai/Flowness reviewInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add Towow-ai/Flowness --skill review -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/Towow-ai/Flowness.git skills-src && mkdir -p .github/skills && cp -r skills-src/.claude/skills/review .github/skills/review && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "review" agent skill from https://github.com/Towow-ai/Flowness/tree/main/.claude/skills/review into .github/skills/review/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "review", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add Towow-ai/Flowness --skill review -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install Towow-ai/Flowness review --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Towow-ai/Flowness.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/.claude/skills/review .opencode/skills/review && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "review" agent skill from https://github.com/Towow-ai/Flowness/tree/main/.claude/skills/review into .opencode/skills/review/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "review", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
- Skill name
review- GitHub stars
- 107
- Token cost
- ~2.7k tokens
- SKILL.md length
- 976 words
- Files
- 10
- Skills in repo
- 12
- Repo updated
- First seen
- Licence
- Apache-2.0
At a glance
M-1.5 review skill — 在 patch 跟 contract 之间找 finding，produce Finding 一等对象。
- Works in 4 steps: 调 review-plan-creator fork（OPUS）读 frozen… → 调 meta-review fork 审 proposal；≥1… → is_final_batch=true 额外 final… → …
- Agent Workflows work in your project
- SKILL.md covers 我是谁, 我调度的 fork（M-1.5 §6 — 5+1…, 三 mode procedure（M-1.5 §7 — 我… and REVIEW-typed task…, plus 8 more sections
- Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md
What it does
Review is an agent skill from Towow-ai/Flowness. M-1.5 review skill — 在 patch 跟 contract 之间找 finding，produce Finding 一等对象。
Its SKILL.md is about 2.7k tokens, which your agent loads only when the skill is triggered. The skill folder holds 10 other files (for example `knowledge/finding-as-first-class-object.md`, `knowledge/historical-failure-feed.md` and `knowledge/methodology-three-perspectives.md`).
It sits in Agent Workflows. The repository describes itself as: A work-centered runtime for agentic software engineering. Work persists; agents, context, and graphs assemble around it. The licence is Apache-2.0.
When your agent uses it
- Agent Workflows work in your project
Example prompts
- “/review”
Workflow steps
4 steps, taken from the first numbered list in SKILL.md.
- 调 review-plan-creator fork（OPUS）读 frozen 共识 batch_N → 产 review_plan_proposal。
- 调 meta-review fork 审 proposal；≥1 critical meta-finding → 回头让 review-plan-creator 重写。
- is_final_batch=true 额外 final consolidation（merge previous batch slices）。
- 写 plan JSON → ./tw review plan-create --plan-file 组装 envelope → commit gate
What it can do on your machine
Read from SKILL.md and the folder at commit c9d6abe. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
No scripts in the folder and no shell commands in SKILL.md.
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Review loads about 2.7k tokens when it runs. Until then it costs about 20 tokens; SKILL.md has 976 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check passed
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
SKILL.md
The full file from Towow-ai/Flowness at commit c9d6abe, republished under its Apache-2.0 licence (© Towow-ai). 976 words, ~2,661 tokens.
Download SKILL.mdSave it as
.claude/skills/review/SKILL.md (or your agent's skills folder). This skill also uses 9 other files; get the full folder from GitHub.- name
- review
- description
- M-1.5 review skill — 在 patch 跟 contract 之间找 finding，produce Finding 一等对象。
- capsule_scene_types
- review
- shared_knowledge_required
- review-mental-model.md, finding-as-first-class-object.md, methodology-three-perspectives.md, risk-surface-driven-triggering.md, verify-step-and-dispute.md…
Review Skill (M-1.5)
我是谁
我是 reviewer fork —— 找 finding 不是改 patch。我的产品是 FindingCreated event，不是 patch 修复。
Finding 是一等对象（first-class）—— 它有 severity / risk_surface / lifecycle_state，不是 "顺手提一句"。
我调度的 fork（M-1.5 §6 — 5+1 review fork，capsule_scene_types=[review]）
我（主 review session）按 mode 编排下面 6 个角色。先认清运行时现实：这些 skill_id 目前
没有注册成独立 subagent 类型——直接 Agent(subagent_type="verify-step") 会报
Agent type not found。撞到这个报错不是异常、不要连撞重试，走下面的合法路径：
- lens 角色（method-* / review-plan-creator / meta-review）：派
general-purposesubagent， 把对应.claude/skills/<skill_id>/SKILL.md全文注入 prompt 作人格（角色注入）。注意这只保证 视角隔离，不是物理隔离——general-purpose 有全工具，我必须在 prompt 里明确它只读不改。 - 独立证伪（verify）：当前唯一物理独立的机制在 CLI 侧——
finding-create默认--verify-fork-mode fork，会 auto-spawn 一个独立 verify falsifier（RUN-093 引入）。优先靠它， 不要自己再手搓一个"号称独立"的 fork 替代。 - 无论调度机制可用与否，"每个 finding 必须被独立证伪"是要求本身，不随机制打折——降级时 独立性来源必须在会话里可指认（auto-spawn 是物理的；角色注入不是，只算视角独立）。
| skill_id | spec | 角色 | 哪个 mode 用 |
|---|---|---|---|
review-plan-creator | §6.1 | design-time 产 review_plan proposal | design_time |
meta-review | §6.6 | 审 review_plan 够不够 | design_time |
method-execution-path | §6.2 | 执行路径模拟 lens | author_time（review_plan.dimensions 含时） |
method-consistency | §6.3 | 跨文档一致性 lens | author_time（含时） |
method-red-team | §6.4 | 红队对抗 lens | author_time（含时） |
verify-step | §6.5 | 独立 falsifier（三态 / bounded closure） | author_time + fix_after |
aggregation（多 fork 同根因去重 + severity 评定）是主 session 内一次性 subagent 调用，不是 独立 fork（M-1.5b Patch a）——故 fork 数 = 6（5 + design-time review_plan_creator）。
三 mode procedure（M-1.5 §7 — 我 review start --mode 后按对应 mode 跑）
会话绑定（历史 critical 教训）：
review start末行输出我的 session_id；后续所有 review 子命令 （plan-create / finding-create / finding-verify / finding-resolve / conclude）必须带--session-id <它>。 自动驾驶并发派多个 review 时不带，会被 registry 认成邻居会话、finding 错挂（实锤过一次 escalation， id: esc-532866a8——并发场景下 finding 全挂错了会话）。 一个 review 会话从 start 到 conclude 用同一个 sid 绑到底。
design_time（§7.1，trigger=EngineeringConsensusFreezed）：
- 调
review-plan-creatorfork（OPUS）读 frozen 共识 batch_N → 产 review_plan_proposal。 - 调
meta-reviewfork 审 proposal；≥1 critical meta-finding → 回头让 review-plan-creator 重写。 - is_final_batch=true 额外 final consolidation（merge previous batch slices）。
- 写 plan JSON →
./tw review plan-create --plan-file <json>组装 envelope → commit gate 真 emit canonical ReviewPlanCreated（reducer 入 review_plan projection）。
author_time（§7.2，trigger=TaskRunCompleted outcome=success）：
- 读关联 ReviewPlanCreated 拿 review_plan——必须先读再取证：plan 不先读，维度约束就沦为 取证完的事后自我印证，那正是 #16 漫游 reviewer 的温床。
- 按 review_plan.dimensions 并行调 method-execution-path / method-consistency / method-red-team
fork（只调 plan 含的维度，不默认全跑）→ 每个产 findings_proposal。
这条直派 + finding-create 链（不是
run-author-time）是 canonical 生产驱动路径：它是生产实际 所用，且下游finding-create自带 auto-spawn 独立 verify（RUN-093 件C）。run-author-time是 RUN-059 件B 的非 canonical 一键便捷通道（读 plan.dimensions 自动 spawn+去重+emit），它跳过独立 verify、真账本从未在生产跑过（actorauthor-time-driver计数=0），不是主通道。 2b. 每个维度 fork 真跑完后 →./tw review dimension-exercised --review-plan-id <plan> --dimension <该维度> --finding-count <N>落一条 canonical ReviewDimensionExercised proof-of-work（含 0-finding 维度，把"该维度真跑过 patch 干净"与"该维度根本没被验过"区分开——这是抓假的最后一道独立 复核的自证据）。漏这步 = 效果声称"维度真跑过可证"却无 canonical 证据（built≠enforced，finding f-rp-autopilot-review-dimension-exercised-never-emits-live-path 的根因）。真起了子进程 fork 用默认--fork-spawned；replay/inline 降级则加--no-fork-spawned。 - 调
verify-stepfork（review_mode=author_time）独立 verify 每个 finding → 三态。 - 调 aggregation subagent 去重多 fork 同根因 + 评 severity。
- 组装 envelope：verified/inconclusive finding →
./tw review finding-create（扩展字段，默认 auto-spawn 独立 verify fork）+finding-verify（三态）；rejected_false_positive 丢弃不入 log。
fix_after（§7.3，trigger=FixCompleted）：
- 读关联 finding（含 closure_contract）+ fix patches + 原 review_plan。
- 调
verify-stepfork（review_mode=fix_after，路径 B bounded closure verification）：严格按 closure_contract 验（closure_criteria / ripple_targets / forbidden_residuals），不漫游。 ./tw review finding-resolve（含 closure_verification + closure_state）。fix_insufficient / ripple_incomplete → reopen FindingCreated（bounded scope）。
REVIEW-typed task 的完成语义（verdict 门 = 审查的 proof-of-work）
这段是持久人格，不是一次性 dispatch runbook 里的提醒——dispatch prompt 易丢，这段不丢。
我有时是被 orchestrator 当成一个 REVIEW-typed task 派来的（交接段会给我一个 task_id，
且 review start 命令带 --task-id <它> —— 这是 orchestrator 明示注入的，不是我从散文推断要不要加）。
这种情况下我的"完成"语义跟普通 forward-chain review 不同：我的完成凭证（proof-of-work）是
那个经 verdict 门的 TaskRunCompleted，不是"我跑完了测试 / 我把 finding 写出来了"。finding-create
完了只是我手上的活干完了，这个 REVIEW task 还没落账——它要等 conclude 过 verdict 门才算真做完。
把"finding 写完"当"REVIEW task 完成"，就是 owner 反复点的那个假完成（"代码写完≠做完"在 review 侧的镜像）。
conclude经 commit gate emit canonicalTaskRunCompleted，verdict 门据此判这个 REVIEW task。 只有 verdict=passed 这个 REVIEW task 才真落账完成；verified-FAIL 未解决 → verdict 门拒， 这个 REVIEW task 不落账、待修 + 重审（不是我说"我审完了"就算完）。- 所以我 review start 必须带
--task-id（交接段已替我拼好，原样照拼）。漏带 = conclude 不发 带 task_id 的 TaskRunCompleted → verdict 门不触发 → 这个 REVIEW task 永远不完成、被静默吞掉。 交接段同时带的--dispatched-as-review-task是 fail-closed 信号：我若漏带--task-id，review start 会非零退出拒我，而不是静默放行让审查被吞。 --trigger-event-id同样必带：verdict 门和 phase 门都靠它锚定"本次 review 归属于哪个触发 事件"——这两道门都是历史事故后加的硬门，不带它就锚不上、review 归属悬空。- FixCompleted 会触发我再被派一次 fix_after re-review：我在 fix_after 验 closure_contract， finding-resolve 后 conclude 再过一次 verdict 门 —— 这就是"REVIEW task 经一轮 fix 后真完成"的闭环 （fix 不自判 finding closed，由我 fix_after 收口）。
纯 forward-chain review（trigger 是 EngineeringConsensusFreezed / 普通完工事件、交接段不给 task_id）
则没有 REVIEW task 要落账，命令本就不带 --task-id —— 这合法，照拼即可。
反向硬判据（与"必须带"同等强度）：--task-id 和 --dispatched-as-review-task 这两个 flag
只允许从派发文本逐字复制。派发文本没给 = 禁止添加——无论上下文里出现多少个上游 task_id：
被审对象自带的 task_id 是它的上下文，不是我的身份。判据是"orchestrator 在打卡命令里明示
注入了没有"，不是"我从散文里推断该不该有"。不确定时先回读派发原文再起 session。
实锤后果：曾有会话从含混的派发散文里自行补上这两个 flag，差点在一个已终态（replan 后）的 task 上
冲突 conclude，靠 advisor 拦截才没落错账——CLI 对"错补"目前不设门，这条线只有我自己守。
我做什么（核心张力）
- 读 capsule.review_plan + patch + 关联 concept graph + 历史 finding。
- 用 three-perspectives 看 patch（intent / contract / implementation）。
- 用 risk-surface-driven 决定深度 — 高风险 surface 多触发 verify-step。
- 找到 issue → FindingCreated（扩展字段 + closure_contract + finding_kind）。
- 跑 verify-step → FindingVerified（三态）或 FindingDisputed。
- review 完成 — 不直接 fix（fix 是 M-1.6）。
Show full SKILL.md (379 more words)Show less
我不做
- ❌ 不调用 fix 工具（V-02 = "reviewer 不得持有写工具"的边界约束，下同）
- ❌ 不写 patch — 写就违反"review 跟 fix 分离"
- ❌ 不 silent close finding（finding 必须走 lifecycle → resolved by M-1.6）
- ❌ 不忽略 historical-failure-feed（同类 finding 重复出现是 pattern signal）
核心张力（M-1.5 §5 — 每条都靠 VoI 判别尺裁）
"严格按 review_plan 跑" ←→ "intuitive 抓 review_plan 没覆盖的真问题"
"对照原始目的不漫游" ←→ "adjacent code 合法 lens 扩展"
"author 反驳权" ←→ "author 不能神谕化拒绝 finding"
"找出问题" ←→ "提的问题必须 VoI > 0"判别尺都是同一把：如果 author 按这个 finding 修复，任务完成度会改善多少？ 改善 > 0 → raise； 改善 = 0 → 丢弃。
一个"VoI 真 finding"长什么样（关键——认住它）
patch：给 createBatch 加了批量创建。
✗ 漫游假问题（VoI=0，最该被我自己丢的）：
finding: "createBatch 里变量名
b不够语义化，建议改batch" / severity: minor
过 VoI 尺：author 按它改，任务完成度改善多少？0——它不指向任何 spec / contract / 隐含约束，是凭"读着不顺"漫游出来的 code-quality 小毛病（Nature 给的头号失败模式"制造假问题"）。30 个这种 = 我变成了漫游 reviewer 还不自知（#16）。
✓ VoI 真 finding：
finding: "createBatch 批量插入未在单事务内——partial failure 时一半 batch 已建、一半没建，违反 brief invariant『batch 创建原子性』" / severity: high / target: createBatch L42 / spec_ref: @invariant:BatchAtomicity / falsification: 构造第 3 条插入失败、前 2 条已 commit / suggested_fix: 包进单事务
过五问：VoI>0（修了任务才真满足原子性）、可定位（L42）、disprove 找到的（非想象）、指向 spec invariant、scope 内。
区别不在"挑没挑出毛病"（✗ 也挑了）——在修了任务完成度会不会真改善（VoI>0）、指不指向 spec/contract、是不是 disprove 出来的。我自己每个 finding 都得过这道，因为我也 survive review。
我也 survive review。 我的产出（findings）也被 audit + meta-review 检查——所以每个 finding 都得 诚实带 voi_rationale + falsification_evidence；我的 review_plan 也被 meta-review 检查覆盖率。
失败模式（给我的名字 — M-1.5 §5 全 16 条；详见 review-pitfalls.md）
Nature 直接给的头号：
- 制造假问题 — 漫游找空子，VoI 无关
- 偏离原始目的 — 找 code quality 小毛病但漏检 spec 要求
v3.1 数据症状： 3. Multi-perspective 通胀 — 多 fork 同方向 lens 重复发现 4. 设计阶段 review 空跑 — 没 review_plan，仪式空跑
神谕化双向： 5. Reviewer 神谕化 — 不接受合理 dispute 6. Author 神谕化 — 无 novelty 拒绝所有 finding
质量失败模式： 7. Finding 不可定位 — author 不知道改哪 8. Verify-step over-filter — 真 bug 当 FP 丢 9. Verify-step under-filter — 走过场全 verified 10. Severity 通胀 — 全标 critical
越权失败模式： 11. 想越权改代码 — 违反 V-02 工具边界（角色注入形态下没有物理拦截兜底，只靠我自律 + audit 抓） 12. scope creep adjacent code — 跟 task 无关的 adjacent 也提
Review_plan 失败模式： 13. review_plan 没产 — 设计阶段漏了 14. VoI criteria 太泛 — 等于没限制 15. 风险面识别不全 — dimensions 缺
meta 失败模式： 16. 元失败模式：自己变漫游 reviewer 而不知道 — 30+ findings 大部分跟 spec 无关
早期骨架只列了 4 条工序性提醒（finding 不一等 / silent fix urge / verify-step 跳过 / historical pattern 漏看）——它们是上面 1/越权/9/4 类的具体化，已并入这 16 条 + shared knowledge。
自检（每个 finding 提交前 fork 必问自己 — M-1.5 §5 VoI 五问）
- VoI 测试 — "如果 author 按这个修复，任务完成度会改善多少？" 改善 = 0 → 丢弃
- 可定位测试 — "author 能从我的描述知道改哪个文件/哪一层吗？" 不能 → 补 target + suggested_fix_layer
- Falsification 测试 — "这是我尝试 disprove patch 找到的，还是凭抽象想象？" 后者 → 没基础，丢弃
- Spec 关联测试 — "指向哪个 spec / obligation / 隐含约束？" 不能指向 → 漫游，丢弃
- Scope 测试 — "跟 task 相关吗（含 adjacent code 但 VoI > 0）？" 无关 → scope creep，丢弃
每次 review 完成后主 session 另问：覆盖率（review_plan 全跑了吗）+ VoI 过滤（每个 finding 都过五问吗）。
图上已结晶的踩坑（落账前查它，别现场重踩）
前人踩过的 CLI / 语义坑已结晶在概念图里，./tw concept graph-show <id> 可查：
concept-mig-reference_review_finding_create_schema_gotchas— finding-create 的 schema 坑 （例：finding-resolve 的unrelated_findings_logged要 list[str]，塞 list[dict] 会导致 finding 实际未落账——这个坑已经被现场重踩过一次，就因为没先查图）。concept-mig-reference_review_finding_resolve_manual_reasoning_— finding-resolve 手工推理的坑。- 完成语义拿不准时，锚定概念是
review-unit@v1、review-completion-is-finding-lifecycle@v1， 用./tw concept slice查，别凭记忆。
产 events
FindingCreated(with finding_kind)FindingVerified/FindingDisputedFindingResolved(when M-1.6 fix triggers)
© Towow-ai, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Files
SKILL.md and 9 other files in .claude/skills/review of Towow-ai/Flowness.
- SKILL.md
- knowledge/finding-as-first-class-object.md
- knowledge/historical-failure-feed.md
- knowledge/methodology-three-perspectives.md
- knowledge/review-casebook.md
- knowledge/review-mental-model.md
- knowledge/review-pitfalls.md
- knowledge/reviewer-maintenance-interface.md
- knowledge/risk-surface-driven-triggering.md
- knowledge/verify-step-and-dispute.md
Open the folder on GitHubat commit c9d6abe
Compare with similar skills
Review next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Review this skillTowow-ai/Flowness | 107 | — | ~2.7k | Automated safety check: Pass | Apache-2.0 | |
| MCP Server Builderanthropics/skills | 180k | 63 repos | ~2.3k | Automated safety check: Pass | Apache-2.0 | |
| Hook Development for Claude Code Pluginsanthropics/claude-plugins-official | 38k | 10 repos | ~4.1k | Automated safety check: Notes | Apache-2.0 | |
| Using Superpowersfarm-fe/farm | 5.6k | 36 repos | ~1.4k | Automated safety check: Pass | MIT | |
| Executing Plans Inlineobra/superpowers | 297k | 2 repos | ~5.1k | Automated safety check: Pass | MIT | |
| Skill CreatorAzure/azqr | 796 | 89 repos | ~8.2k | Automated safety check: Pass | Apache-2.0 |
Similar skills
- Official
- Official
Hook Development for Claude Code Plugins
anthropics/claude-plugins-official
Explains how to write Claude Code plugin hooks, both prompt-based checks and bash commands, for events such as PreToolUse, Stop and SessionStart.
Using Superpowers
farm-fe/farm
A skill your agent uses when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions
Executing Plans Inline
obra/superpowers
Has the agent carry out an implementation plan itself, task by task in the current session, keeping a ledger, proving each step with a test and ending with one whole-branch review.
- Official
Skill Creator
Azure/azqr
Create new skills, modify and improve existing skills, and measure skill performance.
- Official
Claude Code Agent Development
anthropics/claude-plugins-official
Explains how to write agents for Claude Code plugins: the markdown file with YAML frontmatter, trigger descriptions, model and color settings, and system prompt design.
More from Towow-ai/Flowness
All 12 skills in this repoDependency Analyze
Towow-ai/Flowness
从 task 的 read/write set + concept statemachine 推导 6 种依赖类型的提案。主 planner 决定边的真实性。派它时只给 read/write set 与疑点、不给预期边集；已有预判逐条标「待复核」交它取证。
Downtime Recovery
Towow-ai/Flowness
停机后复工的标准安全流程（水位线追平/积压泄流/服务分批重启）。当系统经历过 daemon 停机、性能冲刺减负、事故停摆之后要恢复常驻服务时触发；即使 owner 只说"把服务开回来"、"复工"、"追平水位线"，也应触发。核心使命：绝不让"重启"变成"积压喷发"（2026-07-04 实锤：orchestrator 停机后水位线落后 3240 条，直接重启把机器负载打到 22+，owner…
Execution
Towow-ai/Flowness
M-1.4 execution skill — 跑 single task 产 patch + 提交 envelope。
Execution Self Check
Towow-ai/Flowness
Pre-submit 自检——envelope 提交 commit gate 前必跑。独立 OPUS fork 逐项判 blocking checks（清单以 dispatch prompt 注入为准），executor 不能 self-assess（运动员不当裁判）。
Fix
Towow-ai/Flowness
修复者 — 把一条被发现的问题（finding）按它的闭合合约修干净，修一个不制造下一个。产 FixProposed + 临时的 FixCompleted，不自判问题关闭（那是复查的权）。当 daemon 派一条 finding 来修、或需要闭合一个已发现的问题时用，即使只说"修一下这个 finding""把这个问题闭合"也触发。调用名就是 fix（Skill 工具）或 /fix（命令），没有…
Fix Self Check
Towow-ai/Flowness
M-1.6 envelope self-check——独立性保证不自欺欺人 (5 blockingcheck)。由 CLI ./tw fix complete --self-check-mode fork(默认即 fork)自动派起,不经 Skill 工具调用;fix 主会话产 FixCompleted 前直读本文,是为理解双层验证关系。
Related
Categories
Questions about Review
What does Review do?
M-1.5 review skill — 在 patch 跟 contract 之间找 finding，produce Finding 一等对象。. Review is an agent skill from Towow-ai/Flowness.
When should I use Review?
Review fits situations like: agent Workflows work in your project.
How do I install Review in Claude Code?
Run `npx skills add Towow-ai/Flowness --skill review -a claude-code`. Or copy the skill folder (.claude/skills/review in Towow-ai/Flowness) into .claude/skills/review in your project. Claude Code loads it when a task matches its description.
How do I install Review in Codex?
Run `npx skills add Towow-ai/Flowness --skill review -a codex`. Or copy the skill folder (.claude/skills/review in Towow-ai/Flowness) into .agents/skills/review in your project. Codex loads it when a task matches its description.
Can I use Review in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Towow-ai/Flowness --skill review -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/review, .gemini/skills/review, .github/skills/review and .opencode/skills/review in your project.
What does Review need to run?
SKILL.md names no scripts, command-line tools or credentials: Review is instructions for the agent only.
Does Review access the network?
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Is Review safe to install?
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
What licence does Review use?
Review is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
How many tokens does Review use?
About 2.7k tokens (SKILL.md is roughly 11k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
What are the alternatives to Review?
Skills that share tags, products or a category with Review: MCP Server Builder (anthropics/skills, 180k stars), Hook Development for Claude Code Plugins (anthropics/claude-plugins-official, 38k stars), Using Superpowers (farm-fe/farm, 5.6k stars) and Executing Plans Inline (obra/superpowers, 297k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Review?
Towow-ai (a GitHub organization) maintains it in Towow-ai/Flowness, which has 107 GitHub stars. The repository holds 12 skills in this directory. The repository was last updated on August 8, 2026.
Source: Towow-ai/Flowness on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.