lihaozheCharlie/the-way-here / knowledge-engine/skills/common/retrieval
common-retrieval
Verified知识库共享检索:绑定资料范围,关键词召回、来源和时间辨析、渐进阅读、双链追溯与证据核验。由查询、预测、了解度和构建 Skill 按需调用;不负责回答格式、未来推演或写入。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.1 KB |
| references/source-policy.json | config | 1.4 KB |
| references/tools.md | markdown | 2.2 KB |
| scripts/evidence_reader.py | script | 5.3 KB |
| scripts/snapshot.mts | other | 1.4 KB |
references/tools.md
View on GitHub# 共享检索工具
## 已有冻结资料
优先使用调用方提供的 `read_knowledge_evidence` 专用工具,或其给定的冻结 `evidence_reader.py`。专用工具固定文件、知识库和版本,模型不能更改它们。
命令行工具参数:`python3 <reader> --file <evidence.json> --knowledge-base <id> --hash <hash> --action <action> --purpose <简短问题>`。路径和文字须作为独立参数或正确引用的 shell 参数传递,勿拼接不可信文字执行。
- `overview --offset 0 --limit 20`:分页目录;若调用方提供 lanes/candidates,也会展示。
- `search --terms <关键词1> <关键词2> --offset 0 --limit 20`:正文及标题/别名 OR 搜索,返回总数、行号和短摘录。无四维词表或预测筛选限制。
- `read --page <页面ID> --start 1 --limit 80`:有行号原文及该范围时间线索,按 totalLines 翻页。
- `neighbors --page <页面ID> --offset 0 --limit 20`:出链、反链及未解析链接;只返回当前快照内节点。
每个 action 是独立参数,例如 `--action search --terms 海外 签证`。limit 范围1—200。片段截断会显式标记,引用前回读。共享工具只读;调用方可在私有运行记录保存有上限的动作摘要,工具不写知识库或推理日志。
## 当前知识库的任务快照
在项目根运行,使用本次已选定的知识库 ID:
```bash
THE_WAY_HERE_KNOWLEDGE_BASE=<id> pnpm --dir studio --filter @the-way-here/server exec tsx ../../../knowledge-engine/skills/common/retrieval/scripts/snapshot.mts
```
它使用 `wiki-core` 的同一套来源、时间、别名和双链实现,将完整资料及共享 profiles 写入系统临时目录,输出 file、reader、knowledgeBaseId、inputHash。之后使用上一节工具;不需要 Studio 服务,不加载预测词表、不生成预测。快照仅供本任务,结束时删除本次输出的临时目录,不删除其他任务的目录。若依赖不可用,可以使用只读文件工具按本 Skill 的同样边界检索,不读取预测私有文件作为普通查询输入。
快照是创建时的资料视图,不能作为实时更新订阅。输出中的语义角色、时间和关联线索仍须回读原文核验。
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |