已核实档案CONF. 80/100官方

Tool FS Search

dsh-tool-fs-search

The **model-facing filesystem discovery tools**—`glob`, `grep`—are backed by a packaged ripgrep binary, not b…

AI 能力插件tooltooltool-fs-search
类型
插件
tool
版本
0.1.1-rc.2
能力包版本
分发
open_source
安装分发方式
资料置信度
80/100
达到公开展示门槛
01

先看结论

基于现有公开资料生成,最后决策仍应核对原始来源。

The **model-facing filesystem discovery tools**—`glob`, `grep`—are backed by a packaged ripgrep binary, not by `ctx.fs` provider methods and not by a system `rg` install. Ordinary Node deployments resolve the platform binary from `@vscode/ripgrep`; a pkg single-file runtime resolves the executable's co-located `-rg` sidecar and falls back to the dependency binary when that sidecar is absent. Registration is unconditional because both carriers package ripgrep, so there is no load-time availability probe. Each call spawns the resolved binary through the `ctx.subprocess` seam with a fixed argv vector (`--no-config` prepended so a host `RIPGREP_CONFIG_PATH` cannot inject a `--pre` preprocessor into the unconfined spawn; model-controlled values are plain argv elements — no shell layer exists, so no quoting applies), parses the raw `rg` output, and returns a workdir-relative canonical value. The package injects `tools`, `systemPrompt`, and `subprocess`—deliberately **not** `fs`; `ctx.spillStore` is read opportunistically with `ctx.get()` because formatted-result spill is optional.

02

能力

它能让 Agent / 使用者做什么。

  • 适用 Agent:DeepSeek Harness
  • 适配模型:DeepSeek
03

安装与获取

可复现安装方式。

@deepseek-ai/dsh-tool-fs-search

04

来源与核实

保留来源、时间与置信度,供回溯。

来源
DeepSeek Harness 官方 npm @deepseek-ai/dsh(源码仓库 github.com/deepseek-ai/deepseek-harness)
核实日期
2026-09-04
许可证
MIT
05

实效证据

来源可信不等于任务有效。

暂无通过审核、可复现的同任务 A/B 运行证据;本档案仅证明来源可信,不代表实测有效。

Tool FS Search - 档案 - 效率派 · 效率派