Agent Skills are reusable instruction packs that encode Tradeics developer conventions. When an agent loads a skill, it inherits the right auth headers, API bases, safety rules, and product vocabulary before it writes code or proposes a change.
Skills vs MCP vs llms.txt
| Surface | Role | Load when |
|---|---|---|
| Agent Skills | Procedural guidance (“how to integrate Tradeics”) | Agent starts a Tradeics task |
/llms.txt |
Curated index of docs | Agent needs a map of the hub |
/llms-full.txt |
Full docs corpus | Offline / deep research |
| MCP Servers | Live tools against docs or workspace | Agent must call APIs or search docs live |
Skills never replace authentication: they teach the agent where credentials come from. See Authentication.
Skill catalog (structure)
| Skill id (placeholder) | Audience | Primary outcomes |
|---|---|---|
tradeics-docs |
Any agent answering from official docs | Correct links, no invented routes |
tradeics-platform |
Marketplace / S2P / sales integrators | Correct domain model and workflows |
tradeics-api |
Backend / middleware engineers | REST V2 envelopes, pagination, clients |
tradeics-ai |
Teams using the LLM proxy or AI helpers | Safe inference + assistive boundaries |
Package distribution (npm, GitHub, Cursor marketplace) lands when skills GA. Until then, treat these pages as the canonical outline teams and agents should follow.
Recommended stacking
- Install Tradeics Docs for every Tradeics conversation.
- Add Tradeics API when generating client code or Postman flows.
- Add Tradeics Platform or Tradeics AI for domain-specific work.
- Attach the matching MCP server for live lookups and actions.