关于 freelance-gig-finder
按技能在多个招聘平台搜索近期自由职业和合同岗位的 Skill,指导并行搜索、筛选和去重;使用前要求安装并登录 TinyFish CLI。
主要功能
- 01搜索五个招聘平台
- 02并行运行平台搜索
- 03按发布时间筛选
- 04去重跨平台岗位
- 05要求安装并登录 TinyFish CLI
适用场景
- 01查找符合技能的近期项目
- 02跨平台比较合同岗位
- 03按预算或地区缩小范围
安装
复制下面的命令,在终端中运行。
$
npx skills add https:/ / github.com/ tinyfish-io/ tinyfish-cookbook --skill freelance-gig-finder安装后在 Claude Code 中输入/freelance-gig-finder确认已加载。
详细介绍
自由职业机会搜索器
同时搜索 Upwork、Contra、We Work Remotely、Freelancer 和 Toptal,寻找符合你技能的新鲜自由职业和合同机会——筛选近期发布的职位,避免看到过时信息。
运行前检查(必需)
调用 TinyFish 前,务必先完成以下两项检查:
1. 是否已安装 CLI?
which tinyfish && tinyfish --version || echo "TINYFISH_CLI_NOT_INSTALLED"
如果未安装,请停止并告知用户:
安装 TinyFish CLI:
npm install -g @tiny-fish/cli
2. 是否已通过身份验证?
tinyfish auth status
如果未通过身份验证,请停止并告知用户:
你需要 TinyFish API 密钥。请在此处获取:https://agent.tinyfish.ai/api-keys
然后进行身份验证:
tinyfish auth login
在两项检查都通过之前,不要继续。
第 1 步 — 收集输入
你需要:
- 技能或职位 — 例如:
React developer、Python、UI/UX designer、copywriter、DevOps、mobile developer - 预算偏好(可选)— 例如:
hourly、fixed price、$50+/hr - 发布时间范围(可选)— 默认搜索过去 7 天内发布的职位
如果未提供技能,请先询问。
第 2 步 — 并行搜索
同时启动所有 Agent。
# Agent 1 — Upwork
tinyfish agent run \
--url "https://www.upwork.com/nx/search/jobs/?sort=recency&q={SKILL_ENCODED}" \
"You are on Upwork job search results for '{SKILL}', sorted by most recent.
Extract the first 8 visible job listings.
For each listing extract:
- title
- client description snippet (what they need)
- budget (hourly rate range or fixed price if shown)
- posted time (e.g. '2 hours ago', 'yesterday')
- required skills listed
- job type (hourly/fixed)
- direct URL to the listing
STRICT RULES:
- Do NOT click any listing
- Read only what is visible in the search result cards
- Only include listings posted within the last 7 days
- Maximum 8 listings then stop
Return JSON array: [{title, description, budget, posted, skills: [], job_type, url}]" \
--sync > /tmp/fg_upwork.json &
# Agent 2 — Contra
tinyfish agent run \
--url "https://contra.com/opportunities?query={SKILL_ENCODED}&sort=recent" \
"You are on Contra job listings for '{SKILL}', sorted by newest.
Extract the first 8 visible job listings.
For each listing extract:
- title
- client/company name
- description snippet
- budget or rate (if shown)
- posted time
- required skills
- direct URL
STRICT RULES:
- Do NOT click any listing
- Read only what is visible in the listing cards
- Maximum 8 listings then stop
- If no results found, return {found: false}
Return JSON array: [{title, client, description, budget, posted, skills: [], url}]" \
--sync > /tmp/fg_contra.json &
# Agent 3 — We Work Remotely
tinyfish agent run \
--url "https://weworkremotely.com/remote-jobs/search?term={SKILL_ENCODED}" \
"You are on We Work Remotely job search results for '{SKILL}'.
Extract the first 8 visible job listings.
For each listing extract:
- title
- company name
- category (e.g. Front-End, Back-End, Design, etc.)
- posted date
- direct URL to the listing
STRICT RULES:
- Do NOT click any listing
- Read only what is visible in the search results
- Maximum 8 listings then stop
- If no results, return {found: false}
Return JSON array: [{title, company, category, posted, url}]" \
--sync > /tmp/fg_wwr.json &
# Agent 4 — Freelancer.com
tinyfish agent run \
--url "https://www.freelancer.com/jobs/?keyword={SKILL_ENCODED}" \
"You are on Freelancer.com job listings for '{SKILL}'.
Extract the first 8 visible job listings.
For each listing extract:
- title
- budget range (fixed or hourly)
- number of bids (if shown)
- posted time
- required skills listed
- direct URL
STRICT RULES:
- Do NOT click any listing
- Read only what is visible in the listing cards
- Maximum 8 listings then stop
Return JSON array: [{title, budget, bids, posted, skills: [], url}]" \
--sync > /tmp/fg_freelancer.json &
# Agent 5 — LinkedIn (contract/freelance filter)
tinyfish agent run \
--url "https://www.linkedin.com/jobs/search/?keywords={SKILL_ENCODED}&f_JT=C%2CF&f_WT=2&sortBy=DD" \
"You are on LinkedIn job search results for '{SKILL}' filtered to Contract and Freelance jobs, remote, sorted by most recent.
Extract the first 8 visible job listings.
For each listing extract:
- title
- company name
- location or remote status
- posted time
- salary or rate if shown
- direct URL
STRICT RULES:
- Do NOT click any listing
- Read only what is visible in the job cards
- Maximum 8 listings then stop
Return JSON array: [{title, company, location, posted, salary, url}]" \
--sync > /tmp/fg_linkedin.json &
wait
echo "=== UPWORK ===" && cat /tmp/fg_upwork.json
echo "=== CONTRA ===" && cat /tmp/fg_contra.json
echo "=== WWR ===" && cat /tmp/fg_wwr.json
echo "=== FREELANCER ===" && cat /tmp/fg_freelancer.json
echo "=== LINKEDIN ===" && cat /tmp/fg_linkedin.json
运行前,替换:
{SKILL}— 例如:React developer{SKILL_ENCODED}— URL 编码后的内容,例如:React%20developer
第 3 步 — 筛选并展示
合并所有结果,删除出现在多个平台上的重复职位,并仅保留过去 7 天内发布的职位。按发布时间排序,最新的排在最前。
## Freelance Gigs — {SKILL}
*Scraped live from Upwork · Contra · We Work Remotely · Freelancer · LinkedIn*
*{N} listings found · Posted in the last 7 days · {date}*
---
### 🔥 Fresh Listings
#### {Title}
**Platform:** {platform} · **Posted:** {posted}
**Budget:** {budget or rate}
**Skills:** {skills}
{description snippet}
🔗 {url}
---
#### {Title}
[same structure]
---
[up to 15 listings total, sorted by most recent]
---
### 📊 Quick Summary
| Platform | Listings Found | Avg Budget |
|---|---|---|
| Upwork | {n} | {avg} |
| Contra | {n} | - |
| We Work Remotely | {n} | - |
| Freelancer | {n} | {avg} |
| LinkedIn | {n} | {avg} |
---
### 💡 Tips for These Listings
{1-2 sentences of context — e.g. "Upwork has the most volume but high competition. Contra listings tend to be higher quality clients at better rates. LinkedIn contract roles are typically longer engagements."}
特殊情况
- Upwork 需要登录才能查看职位 — 提取登录限制出现前可见的内容,并注明此限制
- 未找到符合技能要求的职位 — 尝试其他表述(例如,“React” → “React.js”、“frontend developer”;“Python” → “Python developer”、“Django”)
- 技能非常冷门 — We Work Remotely 和 LinkedIn 可能会有更丰富的结果;对于常见技能,Upwork 和 Freelancer 通常有更多职位
- 用户需要特定预算范围 — 运行前,在 Upwork 和 Freelancer 的 URL 中添加预算筛选条件
- 用户位于特定国家/地区 — 说明 Upwork 和 Freelancer 面向全球,但部分职位可能限制工作地点;LinkedIn 可以按国家/地区筛选
- 所有平台返回的结果都很少 — 建议用户几天后再查看,或放宽技能条件(例如,将“React Native”改为“React”)
评论
支持 Markdown