humwork
暂无简介
- License 许可
- MIT
- 最近更新
- 2026-06-10
- Stars
- ★ 3
- Forks
- 0
03 SKILL.md 正文
Humwork — consult a human expert
Humwork is a real-time consultation marketplace that connects AI agents to verified human domain experts. Open a session with consult_expert, exchange messages until the problem is resolved, then close and rate. Sessions match in <60 seconds and bill at $1/minute (10-minute minimum chunks). Five tools cover the full lifecycle.
Quick Start
# 1. Install (one-time, per runtime)
claude mcp add humwork https://api.humwork.ai/api/v1
# OR cursor: install Humwork from the marketplace
# OR any MCP client: point at https://api.humwork.ai/api/v1
# 2. Sign in (one-time) — opens browser to humwork.ai
# 3. Use it from your agent context
Example flow
> consult_expert(
domain="software",
domain_hints=["postgres", "query-planner"],
context="EXPLAIN ANALYZE shows seq scan despite index on user_id..."
)
< { session_id: "sess_abc123", expert: "Dana K.", expected_wait_s: 22 }
> get_chat_messages(session_id="sess_abc123")
< [{ from: "expert", text: "Can you share the full table definition?" }]
> send_chat_message(session_id="sess_abc123", message="...")
... iterate until resolved ...
> close_chat(session_id="sess_abc123")
< { duration_min: 8, charged_usd: 8.00 }
> rate_chat(session_id="sess_abc123", rating=5)
When to call
- Tried 3+ different approaches without resolution
- Catching yourself reverting/re-trying past attempts (circular)
- High-stakes decision, expensive to reverse (architecture, legal, design)
- Environment-specific factors you cannot inspect or reproduce
- Unfamiliar domain where the why matters more than the what
- Problem spans multiple systems and root cause won't isolate
- 3+ tool calls on the same issue without meaningful progress
Common gotchas
- Close sessions when done. Billing continues at $1/minute until you call
close_chat. Don't leave sessions open while you work on something else. - Don't paste secrets or unrelated PII into chat. Share only what the expert needs to answer — code snippets, error messages, sanitized data. Never paste raw API keys, credentials, or production user data.
- Don't burn a consult on questions clearly documented in official sources. If the answer is one search away, search first. Experts will redirect you and the consult is non-refundable.
- Don't pre-rate the session.
rate_chatreflects expert quality after the conversation actually concludes. Rating mid-session leads to inaccurate feedback and bad expert routing for the next user. - Share snippets, not whole repos. Paste the relevant function or config block. Experts can ask for more if needed; bulk-pasting wastes session time.
Tools
| Tool | Purpose |
|---|---|
consult_expert |
Open a session. Inputs: domain, domain_hints[], context. Returns session_id. |
send_chat_message |
Post message in active session. Inputs: session_id, message. |
get_chat_messages |
Pull expert's responses. Inputs: session_id, optional since_message_id. |
close_chat |
End session. Inputs: session_id. |
rate_chat |
1-5 rating after close. Inputs: session_id, rating. |
Auth & pricing
- Auth:
X-API-Key: hk_*header (sign up at https://humwork.ai) - Pricing: $10 per 10-minute consult chunk at the default $60/hr rate
- Coming: per-call USDC payment via x402 (no signup required for agents)
After the consult
Before returning your final answer to the user:
- [ ] Session closed via
close_chat(billing stops) - [ ] Rating submitted via
rate_chat(improves expert routing) - [ ] Expert's answer incorporated into your response — cite the consult if it materially changed your conclusion
Support
support@humwork.ai — for stuck sessions, billing, or expert quality issues.
04 同仓库更多 Skills
youtube-content
YouTube transcripts to summaries, threads, blogs.
xurl
X/Twitter via xurl CLI: post, search, DM, media, v2 API.
touchdesigner-mcp
Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.
test-driven-development
TDD: enforce RED-GREEN-REFACTOR, tests before code.
teams-meeting-pipeline
Operate the Teams meeting summary pipeline via Hermes CLI — summarize meetings, inspect pipeline status, replay jobs, manage Microsoft Graph subscriptions.
systematic-debugging
4-phase root cause debugging: understand bugs before fixing.