跳转至

Capabilities Catalog

Browse production-tested AI building blocks extracted from real enterprise deployments.

All capabilities are available at goldentellus-capabilities.

🔍 What You'll Find

Every capability includes:

  • README: Usage examples, installation, configuration
  • Trust Block: Declared scopes, egress, secrets, cost profile
  • Universal Format(s): MCP Server, Claude Skill, OpenAPI, standalone
  • Production Evidence: Deployed where, for how long, success rate

📦 Current Status

0 capabilities published (first batch in progress)

We're in the early stages of extracting capabilities from our production SNAO deployments. Check back soon or watch the repo for updates.

🧩 Capability Types

Tools (Connectors)

External system integrations: CRM, database, file storage, APIs.

  • Execution: Sandbox (gVisor, Firecracker, Docker)
  • Trust: Requires scopes, egress whitelist, secret refs
  • Formats: OpenAPI (preferred), MCP Server, standalone script

Examples (coming soon): - Salesforce CRM lookup - PostgreSQL query executor - S3 file operations

Skills (Orchestration)

Prompt engineering + RAG for specific tasks.

  • Execution: RAG Adapter (retrieval → rerank → LLM)
  • Trust: Read-only scopes, no egress
  • Formats: Claude Skill manifest, YAML template

Examples (coming soon): - Document summarization - Meeting notes generation - Data analysis prompt chains

Knowledge Packs

Domain-specific knowledge for RAG.

  • Execution: Vector indexing
  • Trust: Read-only, no secrets
  • Formats: Structured docs + schema

Examples (coming soon): - Manufacturing SOPs - Financial compliance rules - Healthcare protocols

External Roles

Customer-facing bot personas.

  • Execution: Gateway Bot runtime
  • Trust: Public context only, outbound filtering
  • Formats: YAML role config

Examples (coming soon): - Lead capture bot - Customer service bot - Partner portal bot

🔐 Understanding Trust Blocks

Every capability declares what it needs:

trust:
  execution_class: executable  # declarative | executable | composite
  required_scopes: ["context:read", "crm:write"]
  egress: ["api.salesforce.com", "*.hubspot.com"]
  secret_refs: ["salesforce_api_key"]
  cost_profile:
    inference_cost: low
    task_tier: non_frontier

Before using: - Check required_scopes: Do you want to grant these permissions? - Check egress: Does it call domains you trust? - Check secret_refs: Do you have these credentials? - Check cost_profile: Does inference cost fit your budget?

🚀 Quick Start

Browse on GitHub

Visit goldentellus-capabilities and explore:

capabilities.md
├── tools/          # External connectors
├── skills/         # Prompt orchestration
├── knowledge/      # Domain knowledge
└── external-roles/ # Customer-facing bots

Install a Capability

MCP Server (Node.js): 

npx @goldentellus/<capability-name>

Claude Skill: 

cp capabilities.mdskills/<name>/manifest.yaml ~/.claude/skills/

OpenAPI Tool: 

# Import spec into your HTTP client
curl -O https://raw.githubusercontent.com/GoldenTellus/goldentellus-capabilities.mdmain/capabilities.mdtools/<name>/openapi.yaml

Standalone: 

git clone https://github.com/GoldenTellus/goldentellus-capabilities
cd capabilities.md<category>/<name>
./run.sh  # or python main.py, node index.js, etc.

🤝 Contribute a Capability

Have a production-tested capability? We'd love to include it!

Requirements: 1. Production deployed (not a demo or experiment) 2. Universal format (MCP / Claude Skill / OpenAPI / standalone) 3. Trust block declared 4. README with usage examples 5. Quality checklist passed

See Contributing Guide for full process.

📚 Learn More

💬 Get Help

Check back soon — first capabilities coming Q4 2026!