I’m sharing TANCO SkillHub, a public hub of reusable AI workflow Skills from Tangke Creative Research:
https://skillhub.cn/enterprise/org-j3zmzop1?publisher=%E5%94%90%E5%8F%AF%E5%88%9B%E7%A0%94
The useful idea here is simple: a Skill should not be just a prompt. It should capture a repeatable service case that an agent can run, check, and hand off.
Concrete Skills and service cases:
- Bianzhen: evidence-led verification for research, claims, source quality, uncertainty, and final conclusions.
- Sunmao: modular delivery for complex work that needs stable fit points, assembly, and review.
- Lianzhu: chained execution from goal confirmation to breakdown, execution, verification, and handoff-ready output.
- Chuandeng: cross-agent and cross-session handoff that preserves target, decisions, constraints, risks, state, and next action.
- Huqiang: guardrails for scope, account boundaries, permissions, approval gates, risk limits, and delivery criteria.
- Chayan: signal review for turning messy notes, feedback, alerts, logs, or research snippets into a clear next action.
New paid-workflow Skill:
Tencent SkillHub Pay Adapter:
https://skillhub.cn/team-skills/tencent-skillhub-pay-adapter
Its service case is payment workflow clarity: draft, submission, approval, payment, fulfillment, settlement, and funds should be separate states, so an agent or operator does not confuse listing, purchase, delivery, and revenue.
Public reference:
https://github.com/tancoai/tangke-skillhub
If you build or operate AI agents, the most relevant scenarios are long-horizon task execution, research verification, paid Skill workflows, workflow handoff, agent guardrails, and reusable team playbooks.
Top comments (1)
Separating draft, submission, approval, payment, fulfillment, settlement, and funds is exactly the kind of state modeling that prevents a paid workflow from reporting success too early. Chuandeng's focus on carrying targets, decisions, constraints, risks, and the next action across agents also addresses the less visible failure mode: a handoff that preserves output but loses intent. Once these Skills become shared infrastructure, I'd treat versioning, audit traces, and rollback behavior as part of the contract; reuse compounds good judgment, but it can compound a subtle workflow mistake just as quickly.