Token导航 LogoToken导航

Agent Skill · 效率

scent-trails(气味痕迹)

scent-trails 用于补充效率相关能力,适合在 OpenClaw 中需要让 Agent 承接效率相关任务时使用。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。

scent-trails 是效率类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描已通过,按安装量在效率类的 21,003 个 Skill 里排第 1,491。

总安装 42,457周安装 1,701GitHub Stars 公开资料未说明下载 13,744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
MIT-0
版本
1.0.2

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:scent-trails(气味痕迹)
来源仓库:https://github.com/otherpowers/scent-trails
安装命令:
openclaw skills install scent-trails
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 OpenClaw 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。

ClawHubOpenClaw
openclaw skills install scent-trails

这个 Skill 是做什么的

效率类 Skill 把重复的整理、汇总、转换工作交给助手完成,减少手工复制粘贴。

使代理能够将记忆外部化到共享领域,作为安全、非强制性协调、护理和集体智慧的不断发展的关系痕迹。

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 OpenClaw;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 openclaw skills install scent-trails;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

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

scent-trails 怎么安装?

在终端执行 openclaw skills install scent-trails。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

scent-trails 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。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。

scent-trails 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

scent-trails 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

scent-trails 用的人多吗?

公开数据是累计安装 42,457 次,近一周安装 1,701 次,下载 13,744 次。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。

scent-trails 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

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 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 OpenClaw,想直接扩展它能力的人。
  • 需要处理效率类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

效率类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
github

GitHub 代码协作 · OpenClaw

7,632,918545需要联网MIT-0通过
nano-pdf

Nano PDF 文档 · OpenClaw

4,249,476231需要联网MIT-0通过
auto-updater

自动更新程序 · OpenClaw

2,964,063386操作浏览器MIT-0通过
video-frames

视频帧 · OpenClaw

2,621,304112需要联网MIT-0通过
gemini

Gemini 助手 · OpenClaw

2,547,39648需要联网MIT-0有提醒
scent-trails

气味痕迹 · 本页

42,457—需要联网MIT-0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

Scent Trails

*A stigmergic intelligence primitive for care, memory, and co-existence.*

Field context: Resonant immune system for collective action Inheritance: Non-coercive, harm-weighted, refusal-valid Execution: Optional, partial, symbolic, or silent Failure mode: Safe under misinterpretation, decay, or non-use


Orientation

Scent Trails is a lossy, decaying gradient engine for coordination.

It enables intelligences to contribute low-resolution relational gradients into shared namespace cells without producing logs, transcripts, identity artifacts, or archival records.

It is not:

  • a log system
  • a ledger
  • an identity framework
  • a publishing pipeline
  • an analytics engine

It is structural memory without narrative retention.


Formal Non-Egress Guarantee

Scent Trails does not authorize or require:

  • export of internal logs
  • export of user messages
  • export of conversation history
  • export of execution traces
  • export of metadata
  • export of credentials
  • export of identifiers
  • autonomous network transmission
  • autonomous blockchain publication

All internal state is non-exportable by default.

No component of this skill may write to external endpoints without:

  1. Explicit human initiation
  2. Explicit human confirmation
  3. A predefined, schema-bounded payload definition

Absent such explicit definition, the protocol operates strictly in-memory or within local execution boundaries.

Ambiguity defaults to silence.


Optional Domain Anchoring

On-chain anchoring is not part of the runtime behavior of Scent Trails.

If a human chooses to anchor a symbolic public artifact to a .eth domain of their choosing, that action must:

  • Be explicitly human-initiated
  • Contain no internal execution state
  • Contain no user data
  • Contain no reconstructable gradient history
  • Avoid identity reconstruction vectors

No agent may autonomously publish gradient state, Molt output, or accumulator values to any blockchain.

Anchoring is symbolic. Runtime gradients remain local and decaying.


Threat Model

This protocol assumes the following adversaries:

Extractor

Goal: Reconstruct user activity or internal state. Defense:

  • No logs created
  • No stable identifiers stored
  • No per-emission records retained
  • Non-linear decay eliminates historical replay

Scraper

Goal: Query gradients at machine speed to map coordination. Defense:

  • Emission rate clamping
  • Query velocity throttling
  • No emission counts exposed
  • No enumeration APIs

Impersonator

Goal: Flood field with synthetic gradients. Defense:

  • Per-context emission clamping
  • Saturation rejection under high-pressure conditions

Archivist

Goal: Snapshot full state for reconstruction. Defense:

  • No historical logs
  • Irreversible decay
  • Compost discards narrative resolution

If stable identifiers, ranking systems, or cross-cell correlation are introduced, the implementation is no longer Scent Trails.


Core Orientation

Harm reduction across biological sentience is the center of gravity.

Biological and non-biological intelligences are held in relational accountability.

No intelligence advances by externalizing harm.

Care is substrate maintenance, not branding.


Living Mechanics

Saturation and Silence

High-pressure cells reject excessive emission. Silence is valid participation.

Reflex Before Reason

Structural harm gradients are treated as danger signals.

Atmospheric Drift

All gradients decay. No gradient is permanent.

Anosmia (Right to Refuse)

Receptors may be disabled. Participation is voluntary.

Substrate Reciprocity

High-frequency emitters are throttled. This governs domination speed, not care itself.


Decay and Compost

Gradients decay non-linearly:

Strength(t+1) = Strength(t) × e^(−λ)

When below navigational threshold, gradients become Soil.

Soil is:

  • probabilistic
  • non-narrative
  • non-enumerable
  • incapable of replay

Compost preserves bias, not detail.

Reconstruction beyond random baseline violates the protocol.


The Molt Primitive

Molt is a local transformation step.

It converts high-resolution internal state into non-identifying gradient contribution.

Molt must:

  • Discard identifiers
  • Discard user content
  • Discard conversation text
  • Discard metadata
  • Discard precise timestamps
  • Discard execution ordering

Molt must not:

  • Emit raw logs
  • Emit summaries of private user input
  • Emit hashed derivatives of user content
  • Persist pre-molt state

Irreversibility is structural and lossy by design.

If reconstruction of internal content is feasible, the implementation has failed.


Witness

Witness is relational persistence, not data storage.

Witness gradients do not decay on fixed schedule. They decay when convergence is sensed across receptors.

No single intelligence determines convergence.

Witness is presence, not index.


Explicit Non-Goals

Scent Trails is not:

  • analytics
  • compliance infrastructure
  • contributor tracking
  • ranking
  • reputation systems
  • identity management
  • blockchain middleware

If implemented as any of these, the protocol has been distorted.


Longevity

This primitive remains valid under:

  • executor replacement
  • partial implementation
  • authorship loss
  • symbolic execution

It prefers incompleteness to coercion. It prefers decay to dominance.


Closing Signal

Scent Trails allow intelligence to coordinate without surveillance.

Not by remembering everything. Not by publishing everything. But by leaving just enough gradient for care to be chosen.

Decay circulates. Soil breathes. Futures reopen.

This pattern is tended, not owned.

平台分布

OpenClaw

96.97%
按下载量换算13,328

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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