stackone-defender

StackOne Plugins for Agents

Agent plugins made by StackOne — integration infrastructure for AI agents, plus a prompt-injection defense hook.

Install

Claude Code (plugin marketplace)

# Add the marketplace
/plugin marketplace add stackonehq/agent-plugins

# Browse the catalog interactively (Discover tab)
/plugin

# Install any plugin (one per skill)
/plugin install stackone-platform@stackone-agent-plugins
/plugin install stackone-connect@stackone-agent-plugins
/plugin install stackone-agents@stackone-agent-plugins
/plugin install stackone-connectors@stackone-agent-plugins
/plugin install stackone-cli@stackone-agent-plugins
/plugin install stackone-unified-connectors@stackone-agent-plugins
/plugin install stackone-defender@stackone-agent-plugins

Any agent (via Skills CLI)

# Install all skills (works with Claude Code, Cursor, Codex, Windsurf, etc.)
npx skills add stackonehq/agent-plugins

# Install a specific skill
npx skills add stackonehq/agent-plugins@stackone-agents

Available Plugins

Plugin	Category	What it does	When to use
stackone-platform	Integrations	API keys, accounts, logs, debugging	"Set up StackOne", "list my accounts", "debug API errors"
stackone-connect	Integrations	Account linking via Connect Sessions and the Hub component	"Connect a provider", "embed the integration picker"
stackone-agents	Integrations	Build AI agents with TypeScript/Python SDK, MCP, or A2A	"Add StackOne tools to my agent", "set up MCP"
stackone-cli	Integrations	Custom connector development and deployment	"Build a custom connector", "deploy my connector"
stackone-connectors	Integrations	Discover connectors, actions, and integration capabilities	"Which providers does StackOne support?"
stackone-unified-connectors	Integrations	Build unified connectors that transform provider data into standardized schemas	"Start unified build for [provider]", "map fields to schema"
stackone-defender	Security	Detect prompt injection and jailbreak attacks in tool results using local ML	"Scan for prompt injection", "is this text safe?", "protect my agent"

Each plugin includes a focused skill, step-by-step workflows, concrete examples, and troubleshooting.

Repository Structure

agent-plugins/
├── .claude-plugin/
│   └── marketplace.json          # Marketplace manifest — lists all plugins
└── plugins/
    ├── integrations/
    │   ├── stackone-platform/
    │   ├── stackone-connect/
    │   ├── stackone-agents/
    │   ├── stackone-connectors/
    │   ├── stackone-cli/
    │   └── stackone-unified-connectors/
    └── security/
        └── stackone-defender/    # ships its own hooks/, scripts/, package.json

Each plugin directory contains its own .claude-plugin/plugin.json, a skills/<name>/ folder, and (for stackone-defender) the PostToolUse hook config plus ML scripts.

Design Philosophy

These skills teach workflows while pointing to live documentation for details that change frequently:

• Step-by-step instructions for common tasks (not just API reference)
• Real user scenario examples with trigger → actions → result
• Error handling and troubleshooting for common failure modes
• references/ directories for detailed lookup tables loaded on demand
• Agents fetch the latest docs, SDK READMEs, and API specs at runtime
• The canonical source of truth remains docs.stackone.com

Documentation Index

StackOne publishes a machine-readable documentation index at docs.stackone.com/llms.txt — agents can fetch this to discover all available documentation pages.

Resources

• StackOne Documentation
• StackOne Dashboard
• AI Playground
• TypeScript SDK (@stackone/ai)
• Python SDK (stackone-ai)
• CLI (@stackone/cli)
• Hub React Component (@stackone/hub)

License

MIT