Agent搜索
- 作者仓库星标 402
- 作者更新于 实时读取
- 作者仓库 affiliate-skills
- 领域
- 通用
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @Affitor · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 即装即用
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: skill-finder
description: > Search and discover Affitor skills by task, stage, keyword, or natural language goal. Returns…
category: 通用
runtime: 无特殊运行时
---
# skill-finder 输出预览
## PART A: 任务判断
- 适用问题:通用任务拆解、检查和交付。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“Stage / When to Use / Input Schema”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于通用任务拆解、检查和交付,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Stage / When to Use / Input Schema”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文没有稳定的斜杠命令要求。安装验证后通常全局生效,直接在对话里点名这个 Skill 并描述任务即可。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件。
先用一个小任务确认它会围绕“Stage / When to Use / Input Schema”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: skill-finder
description: > Search and discover Affitor skills by task, stage, keyword, or natural language goal. Returns…
category: 通用
source: Affitor/affiliate-skills
---
# skill-finder
## 什么时候使用
- 把通用方向的常用动作沉淀成 Agent 可调用的技能 适合处理通用任务拆解、检查、交付和复盘,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代的步骤;通常不需要额外…
- 面向通用任务拆解、检查和交付,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Stage / When to Use / Input Schema」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "skill-finder" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Stage / When to Use / Input Schema
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Skill Finder
Search and discover Affitor skills by task, stage, keyword, or natural language goal. Returns a ranked list of matching skills with descriptions, input requirements, and recommended next steps. Output is a concise Markdown guide.
Stage
S8: Meta — The entry point to the entire Affitor ecosystem. New users don't know what's available. Experienced users forget skill names. Skill Finder bridges the gap — it reads the registry, matches intent to capability, and recommends the fastest path to the user's goal.
When to Use
- User is new to Affitor and asks "what can I do?" or "where do I start?"
- User describes a goal but doesn't name a specific skill
- User wants to find skills by stage (e.g., "what analytics skills exist?")
- User asks "which skill helps with [topic]?"
- User says anything like "find skill", "search skill", "explore skills"
- Chaining: recommended as the first skill for new users before S1-S7
Input Schema
query: string # REQUIRED — natural language: "I want to write a blog review"
# or "what skills help with SEO?" or "analytics skills"
stage_filter: string # OPTIONAL — filter by stage: research | content | blog | landing
# | distribution | analytics | automation | meta
goal: string # OPTIONAL — broader goal: "first commission" | "scale to 1k"
# | "optimize conversions" | "automate my workflow"
Workflow
Step 1: Load Skill Catalog
Read registry.json from the repository root (or from conversation context if already loaded). Parse all skills with their stage, name, slug, and description.
Step 2: Match Query to Skills
Match the user's query against:
- Skill names and slugs (exact match → top priority)
- Skill descriptions (keyword overlap)
- Stage labels and descriptions (if user is browsing by stage)
- Inferred intent (e.g., "SEO" →
seo-audit,affiliate-blog-builder)
If stage_filter is provided, restrict results to that stage.
Step 3: Rank Results
Rank matches by relevance:
- Direct name/slug match
- Description keyword match count
- Stage alignment with user's apparent funnel position
Step 4: Recommend a Path
If the user's goal spans multiple stages, suggest a skill sequence:
- "You want to go from zero to first commission → S1 → S2 → S3 → S5"
- "You want to optimize existing content → S6 (seo-audit, ab-test-generator)"
Step 5: Output Results
Present top 3-5 matching skills with:
- Skill name and stage
- What it does (one sentence)
- What input it needs
- Example invocation prompt
Step 6: Self-Validation
Before presenting output, verify:
- All matched skills exist in the current registry
- Example prompts are copy-paste ready and grammatically correct
- Recommended path follows logical funnel sequence
- Relevance ranking: exact match > partial match > related
- Input needed descriptions match actual skill Input Schemas
If any check fails, fix the output before delivering. Do not flag the checklist to the user — just ensure the output passes.
Output Schema
output_schema_version: "1.0.0" # Semver — bump major on breaking changes
matches:
- skill: string # skill slug
stage: string # e.g., "S6: Analytics"
description: string # one-sentence summary
input_needed: string # what the user needs to provide
example_prompt: string # copy-paste prompt to invoke the skill
relevance: string # "exact" | "high" | "related"
recommended_path:
description: string # why this path
steps:
- order: number
skill: string
action: string # what this step accomplishes
Output Format
- Matching Skills — table with skill name, stage, description, and relevance
- How to Use — for each top match, show the exact prompt to invoke it
- Recommended Path — if the goal spans multiple stages, a numbered sequence
Error Handling
- Empty query: "What are you trying to accomplish? For example: 'write a blog review', 'track conversions', or 'plan a full funnel'."
- No matches found: "No skills match '[query]'. Here are all available stages: [list stages]. Try describing your goal differently."
- Too broad query ("everything"): Show one skill per stage as a sampler, then ask: "Which stage interests you most?"
Examples
Example 1: Specific task query
User: "I want to write a blog review of an AI tool"
Action: Match → affiliate-blog-builder (S3, exact), comparison-post-writer (S3, related), viral-post-writer (S2, related). Show top 3 with example prompts. Recommend: "Start with S1 affiliate-program-search to find the best program, then use S3 affiliate-blog-builder for the review."
Example 2: Stage browsing
User: "What analytics skills are available?"
Action: Filter by analytics stage → show all 4: conversion-tracker, ab-test-generator, performance-report, seo-audit. Describe each with input requirements.
Example 3: Goal-oriented
User: "I'm new to affiliate marketing, where do I start?"
Action: Recommend the beginner path: S1 (affiliate-program-search) → S2 (viral-post-writer) → S3 (affiliate-blog-builder) → S5 (bio-link-deployer). Explain each step in one sentence.
References
registry.json— Machine-readable skill catalog. Read in Step 1.shared/references/flywheel-connections.md— master flywheel connection map
Revenue & Action Plan
Expected Outcomes
- Revenue potential: The fastest path to your first commission is using the right skill at the right time. Skill Finder saves you hours of guessing — it matches your current situation to the exact workflow that generates revenue. Affiliates who follow a structured skill sequence report 3x faster time-to-first-commission
- Benchmark: The typical path to first commission: S1 (find product, 30 min) → S2 (create content, 1 hour) → S5 (distribute, 30 min) = first affiliate link live in 2 hours. First commission typically arrives within 7-14 days
- Key metric to track: Time-to-first-commission. How long from "I started" to "I earned my first dollar"? Use
performance-reportto track ongoing revenue
Do This Right Now (15 min)
- Copy the first recommended prompt from the Recommended Path section and run it immediately
- Don't skip steps — the recommended path is ordered for a reason. S1 before S2, S2 before S5
- Set a goal: earn your first commission within 14 days by running one skill per day
- Bookmark this skill — come back whenever you're unsure what to do next
Track Your Results
After running 3-5 skills in sequence: do you have a live affiliate link? Is it getting clicks? If yes, you're on the path. If no, re-run Skill Finder with a more specific goal ("I have a blog but no traffic" vs "I'm starting from zero").
Next step — run the first skill in your Recommended Path!
Flywheel Connections
Feeds Into
- Any skill —
matched_skillroutes the user to the right skill
Fed By
registry.json— skill catalog with all 44 skills across 8 stages
Feedback Loop
- Track which skills are most frequently requested → surface popular skills higher in recommendations
chain_metadata:
skill_slug: "skill-finder"
stage: "meta"
timestamp: string
suggested_next: [] # Dynamic — depends on matched skill
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核