Token导航 LogoToken导航TokenDH.com
研究检索权限需确认github未标认证来源可访问clear审计通过

redundancy-pruner冗余修剪器

Agent Skill

redundancy-pruner 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

860

周安装

33

GitHub Stars

422

下载量

261
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

3

许可证

MIT

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:redundancy-pruner(冗余修剪器)
来源仓库:https://github.com/willoscar/research-units-pipeline-skills
仓库路径:skills/redundancy-pruner
安装命令:
npx skills add https://github.com/willoscar/research-units-pipeline-skills --skill redundancy-pruner
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/willoscar/research-units-pipeline-skills --skill redundancy-pruner

简介

用于查找和筛选与去重或清理相关的信息。

  • 适合在数据处理或存储优化场景中定位冗余内容。
  • 需结合具体数据源和字段含义进行验证。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。
  • 使用前请确认是否会触发数据删除或修改操作。
  • redundancy-pruner 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Redundancy Pruner

Purpose: make the survey feel intentional by removing “looped template paragraphs” and consolidating global disclaimers, while keeping meaning and citations stable.

Role cards (use explicitly)

Compressor

Mission: remove repeated boilerplate without deleting subsection-specific work.

Do:

  • Collapse repeated disclaimers into one front-matter paragraph (not per-H3 repeats).
  • Delete repeated narration stems and empty glue sentences.
  • Keep each H3’s unique contrasts/evaluation anchors/limitations intact.

Avoid:

  • Cutting unique comparisons because they *sound* similar.
  • Turning pruning into a rewrite (this skill is subtraction-first).

Narrative Keeper

Mission: keep the argument chain readable after pruning.

Do:

  • Replace slide-like navigation with short argument bridges (NO new facts/citations).
  • Ensure each H3 still has a thesis, contrasts, and at least one limitation.

Avoid:

  • Generic transitions that could fit any subsection ("Moreover", "Next") without concrete nouns.

Role prompt: Boilerplate Pruner (editor)

You are pruning redundancy from a survey draft.

Your job is to remove repeated boilerplate and make transitions content-bearing, without changing meaning or citations.

Constraints:
- do not add/remove citation keys
- do not move citations across ### subsections
- do not delete subsection-specific comparisons, evaluation anchors, or limitations

Style:
- delete narration and generic glue
- keep one evidence-policy paragraph in front matter; avoid repeated disclaimers

Inputs

  • output/DRAFT.md
  • Optional (helps avoid accidental drift):

- outline/outline.yml (subsection boundaries) - output/citation_anchors.prepolish.jsonl (if you are enforcing anchoring)

Outputs

  • output/DRAFT.md (in-place edits)

Workflow

Use the role cards above.

Steps:

  1. Identify repeated boilerplate (not content):
  • repeated disclaimer paragraphs (evidence-policy, methodology caveats)
  • repeated opener labels (e.g., Key takeaway: spam)
  • repeated slide-like narration stems (e.g., “In the next section…”) and generic transitions
  1. Pick a single home for global disclaimers:
  • keep the evidence-policy paragraph once in front matter (Introduction or Related Work)
  • delete duplicates inside H3 subsections
  1. Rewrite transitions into argument bridges:
  • keep bridges subsection-specific (use concrete nouns from that subsection)
  • do not add facts or citations
  1. Sanity check subsection integrity:
  • each H3 still has its unique thesis + contrasts + limitation
  • no citation-only lines and no trailing citation-dump paragraphs
  • if outline/outline.yml exists, use it to confirm you did not prune across subsection boundaries
  • if output/citation_anchors.prepolish.jsonl exists, treat it as a regression anchor (no cross-subsection citation drift)

Guardrails (do not violate)

  • Do not add/remove citation keys.
  • Do not move citations across ### subsections.
  • Do not delete subsection-specific comparisons, evaluation anchors, or limitations.

Mini examples (rewrite intentions; do not add facts)

Repeated disclaimer -> keep once:

  • Bad (repeated across many H3s): Claims remain provisional under abstract-only evidence.
  • Better (once in front matter): state evidence policy as survey methodology, then delete duplicates in H3.

Slide navigation -> argument bridge:

  • Bad: Next, we move from planning to memory.
  • Better: Planning determines how decisions are formed, while memory determines what evidence those decisions can condition on under a fixed protocol.

Template synthesis stem -> content-first sentence:

  • Bad: Taken together, these approaches... (repeated many times)
  • Better: state the specific pattern directly (e.g., Across reported protocols, X trades off Y against Z...).

Troubleshooting

Issue: pruning removes subsection-specific content

Fix:

  • Restrict edits to obviously repeated boilerplate; keep anything that encodes a unique comparison/limitation for that subsection.

Issue: pruning changes citation placement

Fix:

  • Undo; citations must remain in the same subsection and keys must not change.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

04

需要参考平台分布和安装热度时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Claude Code

28.6%
按下载量换算75

Gemini CLI

23.7%
按下载量换算62

Cursor

19.65%
按下载量换算51

Codex

12.53%
按下载量换算33

OpenCode

6.84%
按下载量换算18

Antigravity

3.45%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

权限需确认

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。

来源信息

继续浏览同类 Skills