AGENT SKILL

hive-workflow-fusion

Design and run a one-off or reusable HivemindOS workflow from a user's natural-language goal by using hive-capability-search to discover and combine available skills, tools, apps, agents, credentials presence, and delivery channels. Use when the user asks for a workflow, automation plan, "do X using whatever agents/tools we have", or gives a multi-step task that should be executed or rehearsed before being turned into a durable skill.

License 许可
MIT
最近更新
2026-11-10
Stars
★ 13
Forks
4

01 前言 Front Matter

name
hive-workflow-fusion
description
Design and run a one-off or reusable HivemindOS workflow from a user's natural-language goal by using hive-capability-search to discover and combine available skills, tools, apps, agents, credentials presence, and delivery channels. Use when the user asks for a workflow, automation plan, "do X using whatever agents/tools we have", or gives a multi-step task that should be executed or rehearsed before being turned into a durable skill.

03 SKILL.md 正文

Hive Workflow Fusion

Use this skill to orchestrate a user's goal with the best currently available HivemindOS parts. It may produce a one-off run plan, execute a safe rehearsal, or become input for hive-skill-fusion or hive-aeon-fusion.

Capability Discovery

Use hive-capability-search before choosing operators or planning the run. Treat its capability map as the source of truth for selected parts, alternates, availability proof, credential-presence checks, side effects, gaps, and required clarifications.

Do not duplicate broad discovery here. If discovery would repeat, use the existing capability map and move to workflow planning.

Workflow

  1. Parse the intent

    • Break the request into atomic jobs: gather, transform, create, verify, deliver, schedule, or monitor.
    • Identify external side effects such as posting, sending messages, spending money, changing infrastructure, or publishing files.
  2. Retrieve the parts

    • Run or consume hive-capability-search for the original request and atomic jobs.
    • Preserve the returned selected components, alternates, proof, side-effect gates, and gaps.
    • Query fresh app/status surfaces only when the capability map marks an app or external service as selected but not yet reachable.
  3. Choose the operator

    • Prefer a single configured specialist agent when it clearly covers the whole job.
    • Otherwise assign subtasks to the best specialists and make one coordinator responsible for state, approvals, and final delivery.
    • Ask a short clarification only when multiple viable choices are materially different and no policy/default makes the answer obvious.
  4. Plan the run

    • Produce an execution graph with inputs, selected components, viable alternates, outputs, and confirmations.
    • Put read-only discovery and drafting before side-effectful actions.
    • Add checkpoints for user approval before publish/send/trade/deploy actions unless the user has pre-approved that exact class of action.
  5. Execute or package

    • If the user wants action now, run the workflow and capture artifacts.
    • For generated media, verify real artifacts before claiming success: readable file, expected dimensions/duration, nonblank/nonempty content, and a path or delivery reference the user can actually use.
    • For external delivery, do not infer success from an intent, prepared file, target name, or exit code alone. Capture the platform/tool result and require concrete evidence such as success: true, a returned message_id, upload URL, transaction id, post id, or equivalent provider receipt before reporting "sent" or "delivered".
    • If the user wants reuse, hand the capability map to hive-skill-fusion.
    • If the user wants background duty, hand the workflow to hive-aeon-fusion.
    • If a runtime cannot execute or write safely, return the workflow graph and dry-run evidence instead of spinning on discovery.

Output

Return:

  • selected components and why
  • execution graph
  • clarifications or inferred defaults
  • produced artifacts or dry-run outputs
  • blocked/missing components
  • recommended durable skill or AEON workflow, if useful

04 同仓库更多 Skills

LiamVisionary
LiamVisionary/hivemindos

tailscale-generation-skill-authoring

This skill should be used when the user asks to "create a Tailscale generation skill", "add a local generation skill", "document a generation endpoint", "add an image generation provider", "add a TTS provider", "add video generation", "add 3D generation", or needs guidance on where HivemindOS Tailscale or local generation providers belong in the shared skills shelf.

★ 13
LiamVisionary
LiamVisionary/hivemindos

localtts

This skill should be used when the user asks to "generate local TTS", "use local text to speech", "use remote TTS over Tailscale", "make speech audio", "add a TTS provider", "use Qwen3-TTS", or needs TTS endpoint setup through local or HivemindOS Tailscale routes.

★ 13
LiamVisionary
LiamVisionary/hivemindos

karpathy-guidelines

Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.

★ 13
LiamVisionary
LiamVisionary/hivemindos

comfyui-image-generation

This skill should be used when the user asks to "generate images", "generate images with ComfyUI", "use local ComfyUI", "use remote ComfyUI", "use Z-Image", "add an image generation provider", or needs image-generation endpoint setup through local or HivemindOS Tailscale routes.

★ 13
LiamVisionary
LiamVisionary/hivemindos

wrangler

Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

★ 13
LiamVisionary
LiamVisionary/hivemindos

workers-best-practices

Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

★ 13