Token导航 LogoToken导航

Agent Skill · 开发

audience-impact(观众影响力)

audience-impact 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。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。

audience-impact 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,第三方扫描已通过,按安装量在开发类的 21,746 个 Skill 里排第 16,648。

总安装 485周安装 20GitHub Stars 1,102下载 158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:audience-impact(观众影响力)
来源仓库:https://github.com/inkeep/agents
仓库路径:skills/audience-impact
安装命令:
npx skills add https://github.com/inkeep/agents --skill audience-impact
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/inkeep/agents --skill audience-impact

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

audience-impact 提供变更影响的传播模型,帮助评估改动对不同用户群体的影响范围。适用于系统更新、功能发布前的风险评估,识别潜在的用户体验断层。基于角色与部署模式判断影响速度与可见性,标记静默影响项。使用时需明确变更类型与涉及模块,对照传播表进行快速筛查。作为初步判断工具,建议配合详细追踪技能进行深度分析。

来源仓库内路径:skills/audience-impact

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/inkeep/agents --skill audience-impact;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

    它需要接触敏感数据,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

audience-impact 怎么安装?

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

audience-impact 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。

audience-impact 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

audience-impact 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。

audience-impact 用的人多吗?

公开数据是累计安装 485 次,近一周安装 20 次,下载 158 次,GitHub 1,102 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。

audience-impact 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。
  • 来源仓库有 1,102 星,维护活跃度相对可查。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 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。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 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。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
audience-impact

观众影响力 · 本页

4851,102接触敏感数据未明确通过

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

Audience Impact

When assessing a change, ask: who is on the other end, and how fast does this reach them?

This skill provides a mental model — not an exhaustive ledger. Use it to reason about which audiences a change affects, then load the catalog skills for detailed surface tracing when needed.

How to use this skill

  1. Identify which roles and deployment modes the change touches (see definitions below)
  2. Look up the change type in the impact propagation table to see how fast and how visibly it reaches each audience
  3. Flag any silent impacts explicitly — these slip through review most often
  4. If deeper surface tracing is needed, Load: product-surface-areas (customer-facing) or internal-surface-areas (internal subsystems)
  5. Use the deliverables table to determine what artifacts the change needs (changeset, docs, migration guide, etc.)

When multiple audiences are affected, address them in pipeline order: immediate impacts first (Contributor), then versioned impacts (Builder — changeset + migration guide), then deployed impacts (Platform User — QA), then infrastructure (Self-hosted — docs). This matches the natural merge → publish → deploy sequence.

Two-axis model

Audiences decompose into two independent axes:

Role — who is the person? Deployment mode — where does the system run?

These are orthogonal. A Builder can be cloud or self-hosted. A Platform User can be cloud or self-hosted. Assess each axis independently.

Roles

Contributor

Develops the framework source code. Clones the repo, runs pnpm setup-dev, writes and tests code, submits PRs. They interact with build tooling, CI/CD, test infrastructure, database migrations, git hooks, changesets, and dev scripts. Changes to internal tooling, schemas, or dev workflows reach them immediately — they're working in the repo.

Representative surfaces: turbo/biome config, vitest setup, drizzle migration workflow, GitHub Actions, pre-commit hooks, .agents/skills/, AGENTS.md.

Builder

Builds agents using published packages. Runs npx create-agents, installs @inkeep/agents-sdk and related packages, codes against the SDK builder API, uses the CLI (inkeep push/pull/dev), reads docs, and references cookbook templates. They interact with published npm contracts, TypeScript types, config file schemas (inkeep.config.ts), API endpoints, and documentation. Changes reach them at the next npm publish or docs deploy — there's a version boundary between their code and ours.

Representative surfaces: SDK builder functions (agent(), project(), functionTool()), CLI commands and flags, defineConfig() schema, published type exports, AI SDK provider, MCP package, OpenAI-compatible chat API, cookbook templates, agents-docs content.

Platform User

Configures agents through the visual builder (manage-ui) without writing code. Creates projects, configures agents and tools via forms, manages API keys, sets up integrations (Slack, GitHub), views traces, runs evaluations, and manages team members and roles. Changes reach them at the next UI deploy — they see whatever the dashboard ships.

Representative surfaces: agent builder canvas, project/agent/tool/credential forms, API key management, traces dashboard, evaluation UI, work-app integration setup, member/role management, login and onboarding flows.

Deployment modes

Cloud

Uses Inkeep's hosted platform. Auth via SSO/Auth0, credentials managed through Nango-hosted OAuth, services deployed on Vercel. No infrastructure to manage. The PUBLIC_IS_INKEEP_CLOUD_DEPLOYMENT flag gates cloud-specific behavior.

Representative surfaces: Auth0/SSO config, Nango-hosted OAuth flows, Vercel deployment pipeline, cloud onboarding/invitation flows, hosted API URLs.

Self-hosted

Deploys via Docker on own infrastructure. Manages databases (Doltgres, Postgres, SpiceDB), env vars (60+), auth setup (Better Auth, JWT keys), observability (OTEL/SigNoz), and applies migrations. Changes to infrastructure config, Docker setup, or env var schemas affect self-hosters directly — they must act on them.

Representative surfaces: docker-compose.yml, .env.example, database migration scripts, SpiceDB schema, OTEL exporter config, deployment docs (AWS/Azure/GCP/Hetzner guides).

Self-hosted is a modifier — it adds infrastructure surfaces on top of whatever role the person has. A self-hosted Builder manages both SDK integration *and* Docker deployment. A self-hosted Platform User manages both the dashboard *and* the underlying infrastructure.

Impact propagation

This is the primary decision aid. Given a change, how fast and how visibly does it reach each audience?

Change typeContributorBuilderPlatform UserSelf-hosted modifier
Internal tooling (biome, turbo, CI, test infra)ImmediateNoneNoneNone
Database schema (manage-schema.ts, runtime-schema.ts)Immediate (run db:migrate)Next publish (if types change)Next deploy (if UI affected)Must run migrations
SDK/core API (exports, types, builder signatures)Immediate (tests break)Next publish — potentially breakingTransitive (if UI consumes affected API)Transitive (if UI consumes affected API)
CLI commands or config schemaNoneNext publish — potentially breakingNoneNone
Manage UI (pages, forms, components)NoneNoneNext deploy — visible immediatelyNext deploy
API routes or response shapesImmediate (snapshot tests)Next publish — potentially breakingNext deploy (if UI consumes it)Next deploy
Validation schemas (Zod rules, allowed values, defaults)Immediate (tests)Silent — previously valid input rejectedSilent — forms may fail unexpectedlySilent — API calls may fail
Docker/env configImmediate (local dev)NoneNoneImmediate — must reconfigure
Env var semantics (same name, changed meaning/format)Immediate (local dev)NoneNoneSilent — system may behave incorrectly
Documentation contentNoneNext docs deploy — may misleadNext docs deployNext docs deploy
Auth/permissions (RBAC, cookies, tokens)Immediate (tests)Next publish (if SDK auth changes)Next deploy — login/access may breakMust update auth config
Telemetry contracts (span names, OTEL attributes)Immediate (if tests assert spans)NoneNoneSilent — dashboards/alerts may break
Trigger/webhook contracts (schemas, signatures, payloads)Immediate (tests)Next publish — potentially breakingNext deploy — trigger config may breakNext deploy

Impact latency key:

  • Immediate: affects them in their current working session
  • Next publish: affects them when they npm update — version boundary provides a buffer
  • Next deploy: affects them when the UI/API redeploys — no version boundary, but there's a deploy gate
  • Transitive: not directly affected, but may be affected if a downstream surface they use consumes the changed surface
  • Silent: affects them without obvious signal — most dangerous. The system appears to work but behaves incorrectly (e.g., a validation change that silently rejects previously valid input, a telemetry contract change that breaks downstream dashboards)

What counts as "breaking" per role

  • Contributor: CI fails, tests break, dev setup stops working. Loud and immediate — usually caught before merge.
  • Builder: Published API changes (removed exports, changed type shapes, new required config fields, altered CLI flags). Versioned via semver + changesets, but breaking changes in patch versions are especially harmful.
  • Platform User: UI behavior changes (forms that worked differently, removed features, changed navigation). No version boundary — they experience it the moment it deploys.
  • Self-hosted: Infrastructure contract changes (new required env vars, changed Docker config, new database migration steps). Silent if not documented — the system may appear to work but behave incorrectly.

Deeper surface tracing

This skill tells you who cares. For detailed surface enumeration, load the catalog skills:

  • Load: product-surface-areas — customer-facing surfaces (APIs, SDKs, CLI, UI, docs, templates). Maps primarily to Builder and Platform User roles.
  • Load: internal-surface-areas — internal subsystems (build, CI, test, DB, auth, runtime). Maps primarily to Contributor role and the self-hosted modifier.

Deciding what deliverables a change needs

Affected audienceLikely deliverables
ContributorUpdate/add tests. Update AGENTS.md if workflow changes. Update skills if conventions change.
BuilderChangeset (semver). Docs update. Migration guide if breaking.
Platform UserQA the UI change. Update visual-builder docs if behavior changes.
Self-hostedUpdate .env.example. Update deployment docs. Migration guide if infra changes.
Cloud-onlyVerify feature flag gating. Update cloud-specific onboarding if affected.

平台分布

Codex

38.79%
按下载量换算61

Claude

29.3%
按下载量换算46

Cursor

17.73%
按下载量换算28

Gemini CLI

8.78%
按下载量换算14

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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