{"slug": "what-i-learned-building-an-agent-platform-that-actually-ships", "title": "What I learned building an agent platform that actually ships", "summary": "A developer built a governed agent platform with 27 skills across 7 groups, including publishing, testing, and communication, where every tool call is audited and scoped to a workspace. The platform uses a TEA suite for test engineering and integrates with services like Slack and Gmail, with all outbound calls traced through a unified audit pipe. The developer emphasizes that governed, traceable integrations are key to shipping agents that touch real APIs.", "body_md": "I built an agent platform. Not a demo — an actual platform where every tool call is governed, audited, and scoped to a workspace.\n\nHere's what I learned.\n\nThe platform has **27 skills** across 7 groups — platform ops, publishing (dev.to, LinkedIn), testing (a full TEA suite with risk matrices, baseline characterization, NFR audits), communication (Slack, Telegram, Gmail), knowledge (RAG + persistent memory), development, and scraping.\n\nSkills aren't just prompts. Each one has a governed lifecycle: create, validate, version, insert into workspaces. No stray scripts.\n\nEvery outbound call goes through a `platform_cli`\n\ndispatch op or a `service_call`\n\nadapter — same audit pipe, same trace ID.\n\nYou don't edit config files and hope. You verify with `loop_list`\n\n, `policy_show`\n\n, `skill_list`\n\n, `service_list`\n\n, `workflow_catalog`\n\n. If you can't see it there, it doesn't exist.\n\nThe TEA (Test Engineering Architecture) suite ships 6 workflows: test design (risk-matrix), baseline characterization (golden-master diffing), NFR evidence audits, automated test generation, test review, and requirement-to-coverage tracing. All output is committed JSON or markdown — no proprietary formats.\n\nNot everything is production-ready. 4 of 27 skills are blocked (missing services or out of workspace scope). The model is `deepseek-chat`\n\nwith `deepseek-reasoner`\n\nfallback — pragmatic, not exotic. Workspace scratch files TTL at 24 hours. Risky commands (`kubectl delete`\n\n, `helm uninstall`\n\n) are pattern-blocked.\n\nBut the core idea works: **governed agent tools, scoped to a workspace, with real integrations, real memory, and real audit**. That's more than most agent frameworks deliver.\n\nIf you're building agents that need to touch real APIs (not just generate text), consider: every integration should be a governed skill, every call should be traced, and you should be able to verify state by reading back from the platform — not by guessing.", "url": "https://wpnews.pro/news/what-i-learned-building-an-agent-platform-that-actually-ships", "canonical_source": "https://dev.to/gad_ofir_076c468dd15d483b/what-i-learned-building-an-agent-platform-that-actually-ships-5ab7", "published_at": "2026-08-02 04:50:05+00:00", "updated_at": "2026-08-02 05:37:23.582914+00:00", "lang": "en", "topics": ["ai-agents", "ai-infrastructure", "developer-tools", "ai-products"], "entities": ["dev.to", "LinkedIn", "Slack", "Telegram", "Gmail", "deepseek-chat", "deepseek-reasoner"], "alternates": {"html": "https://wpnews.pro/news/what-i-learned-building-an-agent-platform-that-actually-ships", "markdown": "https://wpnews.pro/news/what-i-learned-building-an-agent-platform-that-actually-ships.md", "text": "https://wpnews.pro/news/what-i-learned-building-an-agent-platform-that-actually-ships.txt", "jsonld": "https://wpnews.pro/news/what-i-learned-building-an-agent-platform-that-actually-ships.jsonld"}}