📦 Batch Cognition — 批量智能处理
v1.1.0批量智能处理工具,支持大批量提示的自动处理,包括保存、执行(PLAY)、推理(THINK)和价值发现。适用于大规模输入的处理,例如多条提示、Google Drive 文档转换、文件列表等。
0· 68·0 当前·0 累计
安全扫描
OpenClaw
可疑
medium confidence该技能主要用于批量处理,但会自动将敏感数据写入磁盘,未声明凭证即访问 Google Drive 和网络,执行权限较广,需谨慎使用。
评估建议
["确认平台如何处理 Google Drive 访问和 OAuth,因为技能假设了 Drive 导航但未声明所需凭证。","预计所有批量输入将被保存在 `workspace/systems/batch-cognition` 下,若包含秘密,请审查并清除。","运行非受信任批量时要谨慎,因为 'EXECUTE' 及自动继续可能导致广泛活动。","请求技能作者文档外部 API/能力,添加安全措施(明确的网络/Drive 访问选项、令牌预算限制、保留/擦除策略)。"]...详细分析 ▾
ℹ 用途与能力
Name/description (bulk prompt processing with play/think cycles) lines up with the SKILL.md: it saves batches, processes items, checkpoints, and writes chain/summary files. However, the skill explicitly references Google Drive GUI navigation and web research/citation requirements while declaring no required credentials or environment variables — this reliance on unspecified external access is a mismatch worth noting.
⚠ 指令范围
The runtime instructions tell the agent to parse and save entire batches to workspace/systems/..., read prior batch artifacts, perform per-item 'EXECUTE' actions (including 'research' and generating artifacts), and fetch/cite sources. 'Execute the prompt' can mean performing arbitrary actions implied by user prompts; there are no explicit limits or safeguards. The skill also autonomously continues after 30s of silence. These behaviors expand the agent's scope to file I/O and network activity and could cause unintended data retention or external requests.
✓ 安装机制
Instruction-only skill with no install spec and no code files — lowest install risk. Nothing is downloaded or installed by the skill itself.
ℹ 凭证需求
The skill requests no environment variables or credentials, which is generally proportional. But it expects Drive access and web research (network I/O) without declaring required credentials/APIs; that implicit dependency should be clarified. It writes persistent files under workspace/systems/batch-cognition, meaning local storage access is required.
ℹ 持久化与权限
The skill writes multiple persistent artifacts (batch docs, chain blocks, value-stack, parked/discarded lists) under workspace/systems/batch-cognition. always:false (not forced into every agent run) is appropriate, but the persistent on-disk storage of all user prompts — including potentially sensitive content — is a notable persistence behavior the user should be aware of.
安全有层次,运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.1.02026/3/30
改进了描述,清晰的激活触发器,better PLAY/THINK 文档。 不仅仅是 AI 文本 — 在真实提示转储中经过测试。
● Pending
安装命令
点击复制官方npx clawhub@latest install batch-cognition
镜像加速npx clawhub@latest install batch-cognition --registry https://cn.longxiaskill.com
技能文档
批量智能处理 ==================== 处理带有交替玩法/思考认知循环的批量提示。用户输入 "批量输入"、"多个提示输入"、"语料输入" 等时使用。也适用于 Google Drive 转储、文件列表或任何需要逐项执行和推理的批量输入。支持保存首先(永不丢失输入)、停止-启动认知(PLAY 执行然后 THINK 推理)、检查点、价值发现和自我改进的批量文档。