strongly-ai/strongly-ai-skill

0 stars 质量 61 趋势 20MIT

Claude Agent Skill for using the Strongly.AI platform via its REST API (apps, agents, workflows, MLOps, AI Gateway, and more).

概览

A single Agent Skill that gives Claude everything it needs to build on Strongly through the REST API: deploy apps behind the proxy, provision addons and data sources, run agents and workflows, and operate MLOps, the model registry, and the AI Gateway. Strongly is a platform for deploying apps, data, models, agents, and workflows on managed Kubernetes. Everything you can do in the Strongly UI is also reachable through one REST API at /api/v1. This skill teaches Claude to drive that API correctly, the way an expert operator would: real endpoints only, answers grounded in the resources you actually have, async operations polled to completion, and the proxy, identity, and manifest details that trip most people up handled right the first time. It is written for . It is a how-to, not internal platform documentation.

README

What this is

Strongly is a platform for deploying apps, data, models, agents, and workflows on managed Kubernetes. Everything you can do in the Strongly UI is also reachable through one REST API at /api/v1.

This skill teaches Claude to drive that API correctly, the way an expert operator would: real endpoints only, answers grounded in the resources you actually have, async operations polled to completion, and the proxy, identity, and manifest details that trip most people up handled right the first time.

It is written for you, the Strongly user. It is a how-to, not internal platform documentation.

Who it’s for

Anyone using Claude (Claude Code, Claude in your IDE, or a coding agent) who wants help getting work done on Strongly, whether you are:

building and deploying an app,
wiring up a managed database or an external data source,
standing up an agent or a workflow,
training a model, registering it, and serving it,
or calling third-party and self-hosted models through the AI Gateway.

What’s covered

Area	Reference
Apps: proxy, JWT identity, the manifest, artifacts, deploy via REST	`apps.md`
Addons: managed Postgres, Mongo, Redis, and more	`addons.md`
Data sources: connect external DBs, warehouses, object stores	`datasources.md`
Agents: deploy and chat with Strongly Agents	`agents.md`
Workflows: build and run node graphs, batch and streaming	`workflows.md`
MLOps: AutoML, experiments, fine-tuning, feature store, inference	`mlops.md`
Model Registry: register, version, deploy models	`model-registry.md`
Model evaluation: prediction records, actuals, Record inputs, the model card	`model-evaluation.md`
Drift: baselines, analyses, results, schedule	`drift.md`
AI Gateway: third-party and self-hosted models, keys, guardrails	`ai-gateway.md`
Compute: workspaces, environments, clusters, volumes, code sessions	`compute.md`
Projects: project filesystem and Kanban board	`projects.md`
Governance: policies, gates, evidence, guardrails	`governance.md`
FinOps: costs, budgets, resource groups, schedules	`finops.md`
Marketplace: browse and deploy offerings	`marketplace.md`
Library: memory, rules, prompts, tasks, skills, preferences, pools	`library.md`
Imprints: installable skill + memory bundles	`imprints.md`
A/B testing: compare registry models on live traffic	`ab-testing.md`
Avatars: 3D, portrait, and real-time lip-sync avatars	`avatars.md`
Account, org, and notifications	`account.md`
Python SDK: the `strongly` package, the same API in Python	`python-sdk.md`

How it works

The skill uses progressive disclosure. A small always-loaded SKILL.md teaches the essentials once (how to authenticate, the base URL, the proxy and identity model, and how to find things), then routes to one reference file per feature that is pulled in only when the conversation is about that feature. This keeps the always-on footprint small while giving Claude deep, accurate detail on demand.

strongly/
  SKILL.md              always loaded: auth, base URL, proxy/JWT model, routing
  references/           loaded on demand, one file per feature area
  examples/             runnable snippets

Authentication in two words: it depends where you run

The skill teaches Claude to detect its context automatically:

Inside Strongly (an app, or Claude Code / Codex / OpenCode in a Strongly workspace): the platform URL is in the environment and a bearer token is injected for you. Nothing to configure.
Outside Strongly (your laptop, CI, any external client): you provide your Strongly host and a platform API key (Settings -> API Keys), sent as an X-API-Key header.

Install

Claude Code, copy or symlink the strongly/ directory into your skills folder:

# per project
mkdir -p .claude/skills && cp -r strongly .claude/skills/strongly
# or globally
cp -r strongly ~/.claude/skills/strongly

Then start Claude in that project and ask about Strongly. The skill loads itself.

Other agents (Cursor, Windsurf, Codex, OpenCode), copy strongly/ into that agent’s skills directory (.cursor/skills/, .windsurf/skills/, .agents/skills/, and so on).

In a Strongly workspace, this skill is installed automatically for the coding assistants you enable (Claude Code, Codex, OpenCode), so Claude already knows the platform the moment your workspace starts. Nothing to do.

Requirements

A Strongly account and the host URL of your deployment (for example https://app.strongly.ai), when working from outside Strongly.
A platform API key, created under Settings -> API Keys. This skill never creates keys for you and never handles your passwords.

Contributing

Issues and pull requests are welcome. Keep contributions user-facing (how to use the platform through the API), grounded in real /api/v1 endpoints, and matched to the style of the existing references.

Support

Questions, problems, or ideas: [email protected].

License

MIT. Powered by Strongly.AI.

View this README on GitHub

安装

npx skillfish add strongly-ai/strongly-ai-skill