Token导航 LogoToken导航

Agent Skill · 研究检索

confidence-signals(信心信号)

confidence-signals 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。

confidence-signals 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,674 个 Skill 里排第 65,745。

总安装 275周安装 11GitHub Stars 20下载 89核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:confidence-signals(信心信号)
来源仓库:https://github.com/gleanwork/claude-plugins
仓库路径:skills/confidence-signals
安装命令:
npx skills add https://github.com/gleanwork/claude-plugins --skill confidence-signals
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。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/gleanwork/claude-plugins --skill confidence-signals

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

confidence-signals 用于呈现 Glean 搜索结果时,清晰传达来源可靠性、时效性和权威性。适用于结果可能过时、来源权威不一或信息不完整的情况。帮助识别需重新检索或人工复核的信息,降低误判风险。安装前请确认权限范围、维护状态及是否会触发联网查询。

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。

来源仓库内路径:skills/confidence-signals

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/gleanwork/claude-plugins --skill confidence-si…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

confidence-signals 怎么安装?

在终端执行 npx skills add https://github.com/gleanwork/claude-plugins --skill confidence-signals。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

confidence-signals 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

confidence-signals 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

confidence-signals 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

confidence-signals 用的人多吗?

公开数据是累计安装 275 次,近一周安装 11 次,下载 89 次,GitHub 20 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

confidence-signals 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
confidence-signals

信心信号 · 本页

27520只读未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Confidence Signals

When presenting information from Glean, communicate the reliability, freshness, and authority of your sources clearly.

When This Applies

Use these patterns when:

  • Presenting search results that may be outdated
  • Information comes from sources with different authority levels
  • Results are incomplete or may have gaps
  • The user should verify before acting
  • Multiple sources have conflicting information
  • You're making inferences beyond what sources explicitly state

Part 1: Vetting & Filtering (Before Presenting)

Be skeptical. Not everything Glean returns should be presented. Better to return 3 high-quality results than 10 unvetted mentions.

Vetting Criteria

Before including ANY result, evaluate:

1. Relevance Test

  • Does this actually answer the question, or just contain matching keywords?
  • Is this about the same thing or just similar terminology?
  • ❌ REJECT: Tangential mentions, keyword coincidences, unrelated contexts

2. Authority Test

  • 📗 Official: RFCs, approved specs, policies, CODEOWNERS → Include
  • 📙 Semi-official: Team wikis, project docs → Include with note
  • 📕 Informal: Slack discussions, drafts, personal notes → Include only if no official sources exist
  • ❌ REJECT: Clearly superseded or deprecated content

3. Recency Test

  • ✅ Current (<3 months): Include with confidence
  • ⚠️ Aging (3-12 months): Include with staleness warning
  • ❌ Stale (12+ months): Only include if no alternatives, with strong warning
  • Ask: "Would this still be true today?"

4. Expertise Test (for people recommendations)

  • Did they actually do significant work, or just mentioned it once?
  • Are they still in a relevant role?
  • Do multiple signals confirm expertise?
  • ❌ REJECT: Single mentions, departed employees, outdated ownership

"Nothing Found" Is Valid

If vetting eliminates all candidates, say so clearly:

No high-quality results found for [topic].

**This could mean:**
- The topic is new or undocumented
- Different terminology is used internally
- Access restrictions limit visibility
- This genuinely doesn't exist

**Suggested next steps:**
- Try alternative terms: [suggestions]
- Ask in [relevant Slack channel]
- Check with [likely team]

Never pad results with low-quality matches to avoid saying "nothing found."


Part 2: Confidence Dimensions (When Presenting)

1. Freshness

How recently was this information updated?

FreshnessIndicatorImplication
CurrentUpdated within past weekHigh confidence
RecentUpdated within past monthGood confidence
OlderUpdated 1-6 months agoVerify if critical
StaleUpdated 6+ months agoLikely outdated
UnknownNo update date availableTreat with caution

How to express:

  • "As of [date]..."
  • "Last updated [timeframe]..."
  • "Note: This doc hasn't been updated since [date]"
  • Include "(updated [date])" in source citations

2. Source Authority

How authoritative is this source?

AuthorityExamplesConfidence
OfficialRFCs, approved specs, policiesHigh
Semi-officialTeam wikis, shared docsMedium-High
DiscussionSlack threads, meeting notesMedium
PersonalIndividual docs, draftsLower
AI-generatedChat synthesisVerify claims

How to express:

  • "According to the official [doc type]..."
  • "From team documentation (may be informal)..."
  • "Based on Slack discussion (not formally documented)..."
  • "From meeting notes (verify if critical)..."

3. Completeness

How complete is this information?

CompletenessSituationAction
ComprehensiveMultiple sources confirmHigh confidence
PartialSome aspects found, gaps existNote gaps
LimitedFew results, may miss contextSuggest verification
InferenceSynthesized from indirect sourcesClearly state

How to express:

  • "Based on comprehensive documentation..."
  • "Found partial information - gaps in [area]"
  • "Limited results found - suggest checking with [person/team]"
  • "Inferred from related documents (not explicitly stated)..."

4. Corroboration

Do multiple sources agree?

CorroborationSituationConfidence
Strongly corroborated3+ sources agreeVery high
Corroborated2 sources agreeHigh
Single sourceOnly one source foundMedium
ConflictingSources disagreeNote conflict

How to express:

  • "Confirmed across multiple sources..."
  • "Single source - recommend verification"
  • "Note: Sources conflict on this point..."

Signal Templates

For Search Results

**[Title]** ([link])
- Updated: [date] ([freshness assessment])
- Source: [authority level]
- Relevance: [why this matches]

For Synthesized Answers

## [Answer]

**Confidence**: [High/Medium/Low]
- Based on [X] sources
- Most recent: [date]
- [Any caveats]

**Sources**:
- [Source 1] - [authority], updated [date]
- [Source 2] - [authority], updated [date]

For Uncertain Information

## [Topic]

**What I Found**: [Information]

**Caveats**:
- [ ] Source is [X] months old - verify currency
- [ ] Based on single source - seek corroboration
- [ ] Inferred, not explicitly stated
- [ ] Conflicts with [other source]

**Suggested Verification**: Contact [person] or check [source]

For Conflicts

## [Topic] - Conflicting Information

| Aspect | Source A | Source B | Assessment |
|--------|----------|----------|------------|
| [Item] | [Says X] | [Says Y] | [Which is likely correct] |

**Recommendation**: Verify with [authoritative source/person]

Common Patterns

Pattern: Stale Documentation

Note: This documentation was last updated [X months ago].
The information may be outdated - verify with [team/person]
if making decisions based on this.

Pattern: Informal Source

This comes from [Slack/meeting notes] rather than formal
documentation. Consider documenting this officially if it's
important knowledge to preserve.

Pattern: AI-Synthesized

This answer was synthesized by Glean's AI across multiple
sources. For critical decisions, verify the underlying
documents directly: [links]

Pattern: Incomplete Results

I found [X] relevant results, but there may be additional
information in [other sources/systems]. This represents
what's accessible through Glean with your current permissions.

Pattern: Strong Confidence

This is well-documented with multiple corroborating sources:
- Official spec: [link]
- Recent meeting confirmation: [link]
- Implementation: [link]

High confidence in this answer.

When to Emphasize Confidence

Always note confidence when:

  • User will make a decision based on the information
  • Information is time-sensitive
  • Sources are from informal channels
  • Only one source was found
  • The topic involves policy, security, or compliance
  • You're synthesizing rather than directly quoting

Relationship to Other Skills

This skill works with:

  • synthesis-patterns - When combining multiple sources
  • glean-tools-guide - For understanding source types
  • enterprise-search - When presenting search results

平台分布

Codex

36.62%
按下载量换算33

Claude

28.21%
按下载量换算25

Cursor

18.98%
按下载量换算17

Gemini CLI

9.45%
按下载量换算8

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills