Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

FBS-BookWriter

v2.1.2

FBS 福帮手长文档写作:书/手册/白皮书/长篇报道全流程;Node 脚本驱动 intake、会话恢复、S/P/C/B 质检与 MD/HTML 交付。用户提及写书、出书、章节、大纲、素材、质检、导出、扩写、退出保存时启用。

0· 273·1 current·1 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for duhongchao-fbsir/fbs-bookwriter.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "FBS-BookWriter" (duhongchao-fbsir/fbs-bookwriter) from ClawHub.
Skill page: https://clawhub.ai/duhongchao-fbsir/fbs-bookwriter
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Required binaries: node, npm
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install fbs-bookwriter

ClawHub CLI

Package manager switcher

npx clawhub@latest install fbs-bookwriter
Security Scan
Capability signals
CryptoCan make purchasesRequires OAuth tokenRequires sensitive credentials
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
Name/description (long-form writing, intake/recovery, QC, MD/HTML delivery) match the included Node scripts and CLI usage. Requesting node/npm is appropriate. However the repo also contains extensive host-bridge code (workbuddy/wecom/*, memory adapters, diagnostics, orchestration agents) that go beyond a minimal book-builder; these extras could be legitimate enterprise features but are broader than the brief skill description suggests.
Instruction Scope
SKILL.md explicitly instructs the agent to run many node scripts (intake-router, fbs-cli-bridge, expansion/polish/release governors) and to read/write files under the user's <bookRoot>/.fbs and deliverables directories. That is coherent for a writing tool, but the instructions also reference environment preflight, host diagnostics, and host-memory bridges. The skill cautions against indiscriminate file searches and includes specific safe-read guidance; still, the runtime will execute arbitrary JS in the package and can modify files in the user's bookRoot, so follow the listed constraints closely.
Install Mechanism
No remote download/install spec is included; SKILL.md asks the user to place the skill directory into a known skills path and run npm install in the skill root. That is a standard Node install flow and is proportionate. Because the package already contains hundreds of files and package.json, the risk surface is that 'npm install' will fetch dependencies and then 'node' will run many included scripts — expected but high-impact compared to a small instruction-only skill.
!
Credentials
The skill declares no required environment variables or credentials (primaryEnv none), which is appropriate. However many scripts reference host bridges (workbuddy, wecom client, host memory integration) and runtime hints mention sandbox env injection and optional env flags (e.g., FBS_ALLOW_NO_FOOTNOTE, FBS_BUILD_STRICT_SOURCES). The presence of wecom and host-memory adapter code means the package can call external services if configured; although not requesting secrets up front, it expects host context and may use injected env vars if provided. This expands the attack surface and should be checked if you plan to provide host credentials or enable network access.
Persistence & Privilege
The skill does not request always:true and does not attempt to persist as a platform skill. It does, however, intentionally read/write persistent files under the user's project (.fbs/, releases/, deliverables/) and can back up/modify source files. That is coherent with a writing/CI tool but is a high-privilege filesystem activity relative to a simple read-only helper — treat its file write operations as significant.
What to consider before installing
This package is a large, feature-rich Node-based book authoring system. It will run Node scripts, read and write files under the project/book root (not system-wide by default), and includes optional host-bridge and diagnostics code (WorkBuddy/WeCom/memory adapters). Before installing or running it: 1) review package.json and the top-level scripts you will invoke (e.g., intake-router.mjs, fbs-cli-bridge.mjs, build.mjs); 2) run npm install in an isolated environment (or inspect node_modules before executing scripts) and consider running in a sandboxed container if you are unsure; 3) avoid granting host credentials or injecting sensitive env vars unless you trust the publisher and have audited the wecom/host-bridge code; 4) back up your bookRoot before allowing the skill to modify files; 5) if you need a smaller footprint, ask the publisher for a trimmed package with only build and core writer scripts (no host-bridge/diagnostics) or for provenance/maintainer identity. If you want, I can list the top files to inspect first (scripts that perform IO/network calls) or scan for network endpoints and external host calls in the codebase.
scripts/agents/audit-agent.mjs:83
Shell command execution detected (child_process).
scripts/agents/audit-query-optimization.mjs:10
Shell command execution detected (child_process).
scripts/agents/audit-temporal-accuracy.mjs:10
Shell command execution detected (child_process).
scripts/agents/audit-term-consistency.mjs:10
Shell command execution detected (child_process).
scripts/agents/deploy-agent.mjs:584
Shell command execution detected (child_process).
scripts/agents/research-agent.mjs:109
Shell command execution detected (child_process).
scripts/agents/validation-agent.mjs:75
Shell command execution detected (child_process).
scripts/agents/writing-agent.mjs:234
Shell command execution detected (child_process).
scripts/audit-entry-performance.mjs:37
Shell command execution detected (child_process).
scripts/book-health-snapshot.mjs:52
Shell command execution detected (child_process).
scripts/delivery-chain.mjs:27
Shell command execution detected (child_process).
scripts/env-preflight.mjs:57
Shell command execution detected (child_process).
scripts/evolution-gate.mjs:29
Shell command execution detected (child_process).
scripts/fbs-cli-bridge.mjs:27
Shell command execution detected (child_process).
scripts/fbs-doctor.mjs:25
Shell command execution detected (child_process).
scripts/init-project-memory.mjs:55
Shell command execution detected (child_process).
scripts/intake-router.mjs:118
Shell command execution detected (child_process).
scripts/launch-presentation-preview.mjs:187
Shell command execution detected (child_process).
scripts/lib/channel-pack.mjs:283
Shell command execution detected (child_process).
scripts/lib/git-workspace-changes.mjs:24
Shell command execution detected (child_process).
scripts/lib/pack-skill-gates.mjs:150
Shell command execution detected (child_process).
scripts/lib/workbuddy-host-runtime.mjs:61
Shell command execution detected (child_process).
scripts/midterm-execution-chain.mjs:61
Shell command execution detected (child_process).
scripts/polish-gate.mjs:259
Shell command execution detected (child_process).
scripts/publish-clawhub.mjs:65
Shell command execution detected (child_process).
scripts/quality-audit-incremental.mjs:123
Shell command execution detected (child_process).
scripts/quality-panorama-orchestrator.mjs:198
Shell command execution detected (child_process).
scripts/run-p0-audits.mjs:82
Shell command execution detected (child_process).
scripts/s3-guard.mjs:84
Shell command execution detected (child_process).
scripts/s3-start-gate.mjs:142
Shell command execution detected (child_process).
scripts/smoke-pack-verify.mjs:14
Shell command execution detected (child_process).
scripts/standard-execution-chain.mjs:109
Shell command execution detected (child_process).
scripts/wecom/wecom-client.mjs:97
Shell command execution detected (child_process).
!
scripts/html-delivery-smoke.mjs:168
File read combined with network send (possible exfiltration).
!
scripts/wecom/verify-member.mjs:299
File read combined with network send (possible exfiltration).
Patterns worth reviewing
These patterns may indicate risky behavior. Check the VirusTotal and OpenClaw results above for context-aware analysis before installing.

Like a lobster shell, security has layers — review code before you run it.

Runtime requirements

Binsnode, npm
bookvk97anpbsm38r0sxcxy3yg4rryx84y1x1fbsvk97anpbsm38r0sxcxy3yg4rryx84y1x1latestvk97anpbsm38r0sxcxy3yg4rryx84y1x1longformvk97anpbsm38r0sxcxy3yg4rryx84y1x1manuscriptvk97anpbsm38r0sxcxy3yg4rryx84y1x1openclawvk97anpbsm38r0sxcxy3yg4rryx84y1x1writingvk97anpbsm38r0sxcxy3yg4rryx84y1x1
273downloads
0stars
1versions
Updated 1w ago
v2.1.2
MIT-0

FBS-BookWriter(OpenClaw 封装版)

上游版本:2.1.2(与仓库 FBS-BookWriter 同源)
适配目标OpenClaw 技能模型:SKILL.md + metadata.openclawname 蛇形命名、skillKey 用于 openclaw.jsonskills.entries / 代理 allowlist)。

本目录为 OpenClaw 专用入口:与 WorkBuddy / CodeBuddy 市场通道解耦(发布包内 不含 workbuddy/codebuddy/.codebuddy-plugin/),仍保留完整 references/scripts/scene-packs/assets/package.json,便于在 Gateway 内用 shell/exec 调用同一套 Node 工具链。

认知资产与商业口径(与主仓一致):价值表述统一为「可进化、可分发、可增值」;能力分层统一为「场景包 + 乐包 + 离线/在线会员」。机读见根目录 fbs-runtime-hints.jsoncognitiveAsset;首响 JSON 见 firstResponseContext.cognitiveAssetSnapshot。详细说明见 references/05-ops/cognitive-asset-threeization.md


安装(OpenClaw)

从 ClawHub 安装(推荐,需已发布): 在已登录环境下执行 openclaw skills install fbs-bookwriternpx clawhub@latest install fbs-bookwriter,再于技能根执行 npm install。列表与版本见 ClawHub 上的 fbs-bookwriter

  1. 将技能根目录(文件夹名建议 fbs_bookwriter,与 name 一致)放到任一扫描路径,例如:
    • 工作区:./skills/fbs_bookwriter/(优先级最高)
    • ~/.openclaw/workspace/skills/fbs_bookwriter/
    • ~/.openclaw/skills/fbs_bookwriter/
    • 或在 ~/.openclaw/openclaw.jsonskills.load.extraDirs 中加入父目录
      详见 Skillsskills-configload.watch 默认会热更新技能快照)。
  2. 在技能根执行一次依赖安装:
    npm install(Node ≥18;可选依赖用于 HTML/Docx 等,见 package.json optionalDependencies)。
  3. 重启会话或执行 openclaw gateway restart / 聊天中 /new,用 openclaw skills list 确认加载。

OpenClaw 宿主差异(相对 WorkBuddy)

能力WorkBuddyOpenClaw(本封装)
入口展示intake-router --json + 宿主消费 userFacingOneLiner同样执行脚本拿 JSON;向用户只展示脚本给出的一行摘要 + 最多 3 个选项,勿堆全文
执行命令宿主 bash / 终端使用 OpenClaw exec / shellcwd 必须为技能根(含 package.jsonscripts/
检索宿主 web_search使用 OpenClaw 提供的 联网搜索 / 浏览器(若启用);不可用则按 search-policy 与脚本离线降级
文件工具list_dir / read_file / search_file勿用仅按文件名搜索来探测 .fbs/ 内文件(部分环境不索引点目录);对 .fbs/workbuddy-resume.jsonesm-state.md 等应 read_file 已知路径 或对书稿根 list_dir 确认 .fbs 存在
市场清单workbuddy/channel-manifest.json本包不包含;不影响脚本化写作主链路

执行速查(每次会话)

  1. 统一入口(必做)bookRoot 为用户书稿根目录的绝对路径):

    node scripts/intake-router.mjs --book-root "<bookRoot>" --intent auto --json --enforce-required
    

    需要完整在线场景包时再在命令中加 --full(可能较慢)。 若 JSON 返回 projectAnchor.status=ambiguous,先让用户确认项目根路径,再用确认后的 --book-root 重跑;确认前不要读取任意 .fbs/* 内容。

  2. 恢复:优先 read_file <bookRoot>/.fbs/workbuddy-resume.json;若无则读 chapter-status.md 或进入 S0.5 引导。不要用「全库搜索文件名」代替存在性判断。

  3. 退出:用户说退出时先软询问,再执行:

    node scripts/fbs-cli-bridge.mjs exit -- --book-root "<bookRoot>" --json
    

    回复须包含脚本 JSON 里的 userMessage(会话已记录、下次可继续)。

  4. 阶段与门禁:推进阶段前读 .fbs/esm-state.md;S3.5 扩写须先有用户确认的 .fbs/expansion-plan.md,扩写字数以 node scripts/expansion-word-verify.mjsexpansion-gate.mjs 实测为准(禁止纯模型估算)。 用户说“继续”时按推进优先阈值判断:S0 素材数达到“赛道数×2”、S2 具备章标题+目标字数、S3 已完成≥3章时,优先提议进入下一阶段。 推进 S0→S1 前可运行 node scripts/s0-exit-gate.mjs --book-root "<bookRoot>" --json --confirm-advance 做强制门禁检查。

  5. 写作约束:串行优先;每轮默认最多改 2 个文件;详见 references/05-ops/agent-task-strategy.md。 扩写/精修前先做源文件备份:node scripts/source-write-backup.mjs --book-root "<bookRoot>" --scope expansion --jsonexpansion-gate 默认已自动备份)。 进入 S3.7 精修时优先执行 node scripts/polish-gate.mjs --book-root "<bookRoot>"(先备份再质检),避免绕过门禁直接改稿。 交付收口前执行 node scripts/release-governor.mjs --book-root "<bookRoot>",自动保持唯一终稿并归档旧版本。 对外交付前执行 node scripts/material-marker-governor.mjs --book-root "<bookRoot>" --fix,清理 待核实-MAT[DISCARDED-*] 标注。 全稿/终稿/终审稿发布前必须执行 node scripts/final-manuscript-clean-gate.mjs --book-root "<bookRoot>",若检测到过程标注则不得交付。 任何“已完成/已通过”结论必须附脚本证据(命令或输出路径),禁止仅口头宣称。


意图 → 脚本(节选)

完整表见仓库根目录随包分发的 WorkBuddy 版 SKILL.md 镜像references/01-core/intake-and-routing.md

场景命令
初始化书房node scripts/init-fbs-multiagent-artifacts.mjs --book-root "<bookRoot>"
环境预检node scripts/env-preflight.mjs(参数见脚本 --help;含 glob/iconv-lite 依赖检查)
章节合并 / 健康快照 / 一致性审计merge-chapters.mjsbook-health-snapshot.mjsconsistency-audit.mjs 等(--book-root / --skill-root 指向对应根)
CLI 总线node scripts/fbs-cli-bridge.mjs help

输出格式

  • 对用户:短状态 + 可操作下一步(≤3 条),避免泄露内部规范条文与完整 JSON。
  • 提到术语/文件名时补一句“用途+价值”(例如:chapter-status.md=进度台账,价值是避免漏章和重复写)。
  • 落盘:以 <bookRoot>/.fbs/deliverables/releases/ 为真值;规范见 references/01-core/skill-full-spec.md

错误处理

  • Node/脚本失败:向用户说明失败步骤;若仅缺可选依赖,提示 npm install 或跳过非必需功能。
  • 书稿根未就绪:引导先 init-fbs-multiagent-artifacts 或走 S0 最小集(见 references/01-core/intake-and-routing.md 脚本不可用降级节)。
  • 沙箱无环境变量:若 OpenClaw 在 Docker 沙箱中运行且未注入宿主 env,需在 openclaw.json 中为代理配置 sandbox.docker.env 或关闭沙箱(见官方 skills-config 文档)。

权威文档索引


OpenClaw 封装维护说明:与 pack:openclaw 产物同步;metadata.openclaw.skillKey 与目录名 fbs_bookwriter 应对齐以便 skills.entries 配置。

Comments

Loading comments...