Open-source AI job search: scan job portals, evaluate listings into a structured A-H report with a global 1-5 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)
Upwork freelance skills for Claude Code and Cursor
This repo packages five agent skills for Upwork work: job qualification, proposal writing, invitation replies, profile audits, and client messages. The skills use Upwork guidance plus a Top Rated Plus freelancer playbook, and they expect an optional `upwork-profile.md` for your own niche, links, and rate.
Builders who use an agent to qualify Upwork jobs, write proposals, and handle client conversations.
You can make Upwork work more consistent by reusing the same playbook instead of starting from scratch each time.
What it does
Job qualification
Gives a verdict on whether to apply, skip, or use caution based on client history, red flags, and Connects cost.
Proposal writing
Drafts new proposals or reviews an existing draft using the five-move method.
Invitation responses
Helps you accept, decline, or decline and refer an interview invitation.
Profile optimization
Scores your profile and prioritizes fixes, including the AI-search layer.
Client messaging
Helps with delays, scope creep, rate increases, contract endings, and review asks.
Optional profile file
Reads `upwork-profile.md` so the skills can reuse your identity line, niche, portfolio links, and rate.
How to get it
- 1With the skills CLI
npx skills add abullaisi/upwork-skills --all
README

upwork-skills
English · Bahasa Indonesia
AI agent skills for Upwork freelancers, extracted from a real Top Rated Plus playbook.
Community project. Not affiliated with or endorsed by Upwork.
Five skills that turn Claude, Cursor, or any SKILL.md-compatible agent into a working freelance operator: one that qualifies jobs before you spend Connects, drafts proposals that beat the AI flood, answers invitations fast, audits your profile, and handles the hard client conversations.
Every skill is built from two sources: Upwork's own published guidance (cited inside each skill) and the real, documented playbook of a Top Rated Plus freelancer (top 3%). Where official advice and field results disagree, the skills say so and tell you which one wins. No roleplay personas, no invented stats.
The skills
| Skill | What it does | Reach for it when |
|---|---|---|
| upwork-job-qualifier | Verdict on a job post: apply, skip, or caution. Client-history math, red flags, Connects economics. | "Is this job worth my Connects?" |
| upwork-proposal-writer | Drafts proposals with the five-move method; reviews existing drafts. | "Write a proposal for this job" |
| upwork-invitation-responder | Accept, decline, or decline-and-refer an Invitation to Interview. | "A client invited me, help me reply" |
| upwork-profile-optimizer | Scored profile audit, prioritized fixes, including the AI-search (AEO) layer. | "Why am I not getting invites?" |
| upwork-client-messenger | Mid-contract messages: delays, scope creep, rate increases, contract endings, review asks. | "The client keeps adding scope" |
Together they cover the whole loop: choose well, get chosen, be findable, and run the relationship.
Install
With the skills CLI:
npx skills add abullaisi/upwork-skills --all
Or manually: copy any skill folder into your agent's skills directory (.claude/skills/ for Claude Code) and it loads on the next session.
The profile shortcut
Every skill reads an optional upwork-profile.md from your working directory or ~/.claude/: your identity line, niche, portfolio links, and rate. Store it once and the skills stop asking. Example:
# My freelance profile
- Role: UI/UX designer, 5+ years, SaaS dashboards
- Rate: your hourly rate
- Portfolio: 3-5 links you actually send to clients
Why these instead of just asking an AI
Upwork's built-in AI drafts proposals for every freelancer on the platform, free. That's the problem: generic AI output is now the baseline clients scroll past. These skills encode the parts a platform tool can't or won't: judgment about which jobs to skip (platforms profit when you apply more, not less), a proposal method with the field corrections that won real contracts, and client-conversation moves tested in real threads. The sources are cited in every file, so you can check the receipts.
Contributing
Contributions are welcome, and small ones count:
- Corrections: guidance that's gone stale, a mechanic that changed.
- Translations: any skill, any language. Bahasa Indonesia versions are a first-class goal.
- New skills: more Upwork craft (pricing, portfolio prep, niche workflows) fits here. Other platforms deserve their own pack. Open an issue first so we agree on scope.
Keep the house rules: plain human voice, no invented statistics, sources cited, no client names or real rates in examples, and no scraping or automation that violates any platform's terms.
Freshness
Guidance verified as of July 2026. Upwork changes its mechanics regularly: Connects pricing, badge rules, search behavior, the AI features on both sides of the marketplace. Every skill cites its sources so you can check them yourself, and each one flags where official guidance and field results disagree.
If something here no longer matches what you're seeing on the platform, that's the most valuable issue you can open. Corrections from working freelancers are what keep this accurate, and they are welcome in English or Bahasa Indonesia.
License and disclaimers
MIT. Use it, fork it, ship it.
This project is not affiliated with or endorsed by Upwork. "Upwork" appears descriptively, because that's the platform the guidance is about.
Built by Imam Abullaisi and the Wargi Freelance community (Indonesian freelancer community, 950+ members). The lessons repo this pack grew from is coming next.
Files in the repo
- .github
- assets
- upwork-client-messenger
- upwork-invitation-responder
- upwork-job-qualifier
- upwork-profile-optimizer
- upwork-proposal-writer
- .gitignore
- CODE_OF_CONDUCT.md
- CONTRIBUTING.md
- LICENSE
- README.id.md
- README.md
- SECURITY.md
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More skills
A skill to stop your coding agent from burying the answer. ADHD-friendly output.
De-AI writing skill for any Agent Skills-compatible agent (77+ via the Skills CLI), with native plugins for Claude Code, Codex, Grok Build, and Antigravity. Narrative-architecture repair for fiction, venue-matched rules for professional prose. Based on StoryScope (arXiv:2604.03136).
Academic Research Skills for Claude Code: research → write → review → revise → finalize
Agent skill that removes signs of AI-generated writing from text
AI video skill for Claude Code & Codex — cinematic product videos with Remotion: 152 shot recipe cards, 209 motion previews, a production-ready template