Quite surprised by myself for being consistent at posting almost every week for the past month. This specific post was not planned but inspired by the great work by the team published here.
I really loved the concept of the original skill created by and
. It's available here if you want to give it a try: https://github.com/Mantequilla-Soft/hive-master-skill.
Why a separate version
The original approach is great, but I wanted a few things tuned more to my personal taste and workflow:
- compatibility with skills.sh for installation and easy update
- broad support for OpenCode, Codex and Claude Code (+ others)
- focus the implementation on
hive-txbyas a modern and simple library to interact with the Hive blockchain
So I started from that inspiration and rebuilt the hive-developer skill in my own skills repository: https://github.com/asgarth/skills. Or link to the specific folder.
Easy installation with skills.sh
You can install it via skills.sh tooling from the same repository:
npx skills add https://github.com/asgarth/skills --skill hive-developer
The setup wizard will automatically place it in the correct location for the agents you use (Codex/OpenCode/Claude), depending on your local skill directory conventions.
Support multiple agents
Support for multiple agents is an easy win. And can be achieved with the original skill as well if manually installed in the correct location.
I feel like we should do our best as a community to not focus on Claude Code as our only/main agent going forward. Support for multiple agents and models seems very much aligned with the core proposition of the Hive blockchain and there is no reason to not apply it here.
hive-tx as a modern library for Hive
I wanted a more modern and versatile base library for this skill, and for my use cases hive-tx is the right fit.
Main reasons:
- simple integration and clean code
- actively maintained and improved
- support for the new Hive REST APIs through
callREST(...) - automatic node fallback with multi-node config + retries
callWithQuorum(...)for critical reads where consistency matters- clean transaction flow with optional wait-for-inclusion behavior
That gives better building blocks for robust automation and agent-driven workflows.
What is inside the refreshed skill
The current hive-developer skill covers:
- RPC + REST usage patterns
- operation examples (
vote,comment,transfer,custom_json, etc.) - key/auth mapping (posting vs active)
- reliability defaults (failover, no blind rebroadcast, status handling)
- streaming/realtime patterns via polling and block processing
Here is the skill structure at a high level, with each file mapped to its technical role:
| Layer | File | Technical purpose |
|---|---|---|
| Entry point | SKILL.md | Defines when the skill should trigger, core defaults, workflow, and output guardrails for agents. |
| API | api-reference.md | Documents hive-tx read stack: callRPC, callREST, node config, retries, and callWithQuorum usage. |
| Transactions | operations.md | Provides operation patterns (vote, comment, transfer, custom_json) with signing + broadcast flow. |
| Auth | keys-and-auth.md | Maps operation authority to key type (posting/active), key validation, and signing safety rules. |
| Reliability | reliability.md | Covers failover strategy, wait-for-inclusion behavior, status handling, and safe rebroadcast policy. |
| Realtime | streaming-and-realtime.md | Describes polling/block-processing architecture for near-realtime flows with callRPC. |
Final note
Again, full credits (and a chunk of the rewards on this post) to Mantequilla Soft for the original concept.
This refresh is not a replacement, it is more like a focused flavor adapted to my workflow and preferred stack. If you prefer to work with dhive you should definitely use the original skill.
If you are building AI agent tooling on Hive, feel free to try it and share feedback/PRs.