Claude-Code-Plugins on Start AI Tools - Presented by Intent Solutions

BigQuery Integration, 23 Skills at L4 Quality: Nixtla v1.8.0

Wed, 10 Dec 2025 10:00:00 -0600

26 commits. The day the three plugins stopped being independent projects and started being a pipeline.

Four days of infrastructure, quality enforcement, and security hardening led to this: the moment where separate components become a system. v1.8.0 isn’t the biggest release by commit count. It’s the most important by what it proves — three independently developed plugins can compose into a reliable workflow with tested failure handling.

3-Plugin Workflow Integration

Each plugin had been built and tested in isolation. Search-to-slack finds information. Baseline Lab runs forecasts. BigQuery Forecaster queries production data. All three worked. None of them talked to each other.

21 Skills and a Production Validator: Nixtla v1.6.0

Mon, 08 Dec 2025 10:00:00 -0600

66 commits. The biggest single-day push the Nixtla repo has seen. Three major workstreams running in parallel: skill extraction, validator construction, and TimeGPT lab expansion. Each one would have been a reasonable day on its own.

The commit rate was sustainable because yesterday’s CI optimization eliminated the feedback loop bottleneck. Every push validated in under a minute. That changes behavior — you commit smaller, more frequently, with higher confidence. The 66 commits aren’t 66 features. They’re 66 incremental steps where each one was tested before the next began.

Skills Infrastructure from Scratch: 8 SKILL.md Files, an Installer, and a Compliance Framework

Mon, 01 Dec 2025 10:00:00 -0600

53 files. 13,445 lines. One goal: give the Nixtla workspace a skills layer that doesn’t exist yet.

What Needed Building

The Nixtla workspace had plugins. It had documentation. It had release discipline. What it didn’t have was a way for external tools to discover and invoke its capabilities through a standardized interface.

Skills are that interface. A SKILL.md file declares what a skill does, what inputs it needs, what outputs it produces, and what permissions it requires. An installer package reads those declarations and wires them into Claude Code. A demo project proves the whole chain works end-to-end.

Nixtla Baseline Lab: Real statsforecast Baselines via MCP Server

Mon, 24 Nov 2025 10:00:00 -0600

Yesterday was plugin scaffolding. Today the Baseline Lab gets real models running real data through a real MCP server.

Phase 2: Plugin Manifest and MCP Config

The Baseline Lab plugin needed two pieces of infrastructure before any forecasting code could run.

Plugin manifest. Standard plugin.json declaring the plugin’s name, version, dependencies, and MCP server configuration. The MCP config tells Claude Code how to start the server, what tools it exposes, and what permissions it needs (filesystem read for data files, network for optional Nixtla cloud API calls).

Nixtla Repo: Zero to v0.2.0 with a Working Search-to-Slack Plugin

Sun, 23 Nov 2025 10:00:00 -0600

New repo. Empty directory. By end of day: v0.2.0 tagged, a working plugin shipping, and a GitHub Pages site live.

Repository Initialization

The Nixtla Claude Code plugins workspace needed proper infrastructure from commit one. Not a README and a prayer – actual CI/CD, documentation filing, and GitHub Actions workflows.

First commit: project scaffolding with GitHub Actions for CI, automated doc filing using the standard 000-docs numbering scheme, and a base configuration for the plugin workspace. GitHub Pages site deployed from the repo’s docs output.

Marketplace Security Framework and the False Positive Fight

Mon, 13 Oct 2025 10:00:00 -0500

Three days after launching the Claude Code plugin marketplace, the security story was “trust me, I checked.” That needed to change before anyone installed a community plugin on their machine.

The Security Framework

The day started with a full SECURITY.md — 407 lines covering a 6-vector threat model, three trust tiers (Community, Verified, Featured), responsible disclosure policy, and SLA commitments. Standard structure, but adapted to the plugin marketplace context where users execute community code inside their development environment.