01
Skill authoring and evaluation
Write the trigger, the inputs and outputs, and the tool boundaries. Check normal results and failure paths against fixed cases. You receive a Skill the platform can read.
01 / Delivery studio
HopAgent is a one-person company. We author, evaluate and debug these three kinds of deliverables so major agent platforms can discover and list them.
02 / Services
Delivered to spec means the platform reads it, the task runs, and the entry point is clear.
01
Write the trigger, the inputs and outputs, and the tool boundaries. Check normal results and failure paths against fixed cases. You receive a Skill the platform can read.
02
Tighten the system instructions, tool permissions and handoff. Reproduce failures on a real task until the same request can be run again.
03
Connect through MCP, Actions and bots. The platforms are in the list below. Use only interfaces that already exist. Do not invent APIs.
03 / Method
Each step leaves something you can check. Nothing rests on a verbal explanation.
01
Write the capability boundary, inputs and outputs, the tool list, and the format the target platform asks for.
Output · Spec and tool list
02
Use fixed tasks to check correct results, refusals, timeouts and overreach. Keep a record when something fails.
Output · Evaluation record
03
Change the setup and the connection from the failure record until the same task can be run again reliably.
Output · Re-run log
04 / Packages
Diagnosis, delivery and listing are quoted by scope. This page does not set a price. Write to hopagent@163.com with the platform, the deliverable and where things stand.
See where an existing Skill, Agent or Connector is stuck: format, permissions, call chain, or missing evaluation. You receive a problem list and an order of changes.
Build a callable Skill, Agent or Connector from scratch or from a draft, with evaluation records that can be run again.
After diagnosis or delivery, arrange the checklist, description and entry against the target platform's published requirements. No guarantee of approval, exposure or ranking.
05 / Platforms
These are the platforms we prepare deliverables against. This is not a partner list. We do not write, and we do not claim, any unpublished interface.
Hugging Face Spaces are used only when a deliverable fits a Space for display or calling. WorkBuddy, Qwen Go (千问 Go) and Doubao Work (豆包工作) are agent work platforms. We use what they have already published. API keys stay in your account. This site does not hold them.
06 / FAQ
Listing, keys, replies and pricing — the answers are below.
A prompt is only material. We deliver a Skill, Agent or Connector with boundaries, evaluation records and an entry point, so a platform can discover and list it later.
No. Listing depends on each platform's own review and rules. We prepare material against published requirements and promise no approval, exposure or ranking.
No. We do not hold keys. Connections use your account and permissions. Debugging runs in your environment. If a step needs a borrowed credential, it is temporary and you can revoke it.
The founder replies directly. There is no support queue. After you write, we confirm the scope, then agree on timing.
Diagnosis, delivery and listing are priced by scope. This page does not list a number. Write to hopagent@163.com with the platform, the deliverable and where things stand.
07 / Contact