AI role --MCP Tool & Skill Developer SmartAssist Enterprise Agent Platform — MCP Tool & Skill Development Location - Remote Team / Platform: Serenity / SmartAssist Enterprise Agent Platform Reports to: SmartAssist Platform Lead ____________________________________________________________

Job Description

SmartAssist assembles specialized AI agents from two building blocks: MCP tool servers that expose enterprise systems, and skills that package those tools with business instructions, authorization scope, and workflows. We are engaging a contract MCP Tool & Skill Developer to build and maintain both, providing the governed capabilities that the platform's specialized agents rely on. On the tool side, the role builds single-platform MCP servers (Python / FastAPI), deployed to Kubernetes behind Azure APIM with Entra JWT authentication, created from the platform's MCP template and following its auto-discovery and single-responsibility conventions. On the skill side, the role authors and validates skills — SKILL.md manifests that declare allowed_tools (which also serves as the OPA authorization scope), pre-flight inputs, lifecycle stage, and workflows — which specialized agents consume through the Skills Registry. The two outputs are complementary: well-designed tools plus well-scoped skills are what let the agent-development team construct reliable, least-privilege agents without changing application code. Core Job Functions Build new MCP tool servers from mcp-template as single-platform adapters (one backend per server), following the auto-discovery conventions in src/tools/ and the token-efficient, single-responsibility file structure. Implement, document, and test individual tools (list / get / create-style operations plus a bootstrap guide), with unit tests and integration tests against backend systems. Author skills as SKILL.md manifests — declaring allowed_tools (the OPA authorization scope), pre-flight inputs, natural-language aliases, tags, and lifecycle stage — using the create-skill and audit-skill-config skills. Validate skills locally with validate_skills.py and pass audit-skill-config with no failures before promoting to production; publish via the registry upload scripts. Define and encode skill workflows: execution steps, error-handling tables, and gotchas, so agents can run them reliably. Integrate tools and skills with APIM, the PostgreSQL-backed Skills Registry, and OPA policy enforcement, ensuring least-privilege tool scoping. Deploy and operate servers on Kubernetes (Docker images, Entra JWT auth, health endpoints); maintain certificates and environment configuration. Partner with the AI agent development team to shape tool and skill contracts to real agent needs, and keep the catalog consistent as it scales (~30 servers, ~1,500 tools, 1,000+ skills). Create evaluation suites and benchmark scenarios that verify tool functionality. Align with existing observability and logging capabilities to support agent execution tracing, tool-call analysis, and operational monitoring. Create technical documentation, architecture diagrams, and other knowledge-transfer sufficient for internal teams to maintain delivered components independently. Experience Strong Python engineering experience, including FastAPI or similar; comfortable building and testing HTTP / API services. Experience implementing the Model Context Protocol (MCP) or a comparable tool / plugin server interface; understands tool schema design and clear, LLM-consumable outputs. Solid REST / API integration experience against enterprise backends (e.g., ServiceNow, GitLab, Ansible, Jira, or similar), including authentication and error handling. Containerization and orchestration experience: Docker and Kubernetes; familiarity with API gateways (Azure APIM) and Entra / OAuth2 / JWT auth. Experience with automated testing (pytest), CI/CD, and Git-based multi-repo workflows. Able to write precise, structured configuration and documentation (YAML / Markdown manifests) with strong attention to correctness and consistency. Nice to have: policy-as-code (OPA / Rego), PostgreSQL, skill / prompt design for LLM agents, and experience in regulated or large enterprise environments.