14 min read
The Agent Delivery Harness: An Opinionated Multi-Model Layer
An agent delivery harness turns human intent, experience, and judgment into accountable, evidence-bound decisions across one or more model runtimes.
Writing
Analysis of agentic systems, accountable delivery, and the architecture required to move AI work beyond the prototype.
23 published analyses
Research lanes
Four editorial entry points organize the archive. The complete topic record remains available below.
- Agent systems6Orchestration, durable execution, and the architecture around model-driven work.
- Delivery architecture4The control surfaces that turn experimental capability into maintainable systems.
- Tools and capability10How coding agents, skills, protocols, and platforms change the engineering workflow.
- Security and assurance1Operational boundaries, evidence, and risk controls for AI-enabled delivery.
Latest analysis
A chronological record. The lead essay is held above and does not repeat here.
- 7 min read
7 min read
Compiling Agent Config to Claude Code, Codex, and Copilot
Learn how base compiles one vendor-neutral canon into native agent config for Claude Code, Codex, and GitHub Copilot — with drift protection in CI.
- 8 min read
8 min read
Compile Your Agentic System: base, skillsmith, and ObjectCore
Explore three projects that treat agent config as compiler output: validated sources in, drift-gated artifacts out — base, skillsmith, and ObjectCore.
- 7 min read
7 min read
ObjectCore Deep Dive: A Plugin Marketplace as a Software Factory
Explore how ObjectCore serves one plugin marketplace from git in dev and a database in production — same pure function, SHA-pinned, provenance-attested.
- 8 min read
8 min read
Quality Gates for Claude Code Skills: Inside skillsmith
Discover the pipeline that keeps Claude Code skills honest: 14 quality rules, 7 security rules, trigger evals with a 0.85 floor, and a byte-exact drift gate.
- 10 min read
10 min read
Building a Deep Research Agent That Verifies Its Sources
Learn how I built a deep research agent that fans out across sources, then refutes its own findings with three skeptics before reporting a single cited claim.
- 11 min read
11 min read
Durable Agent Orchestration with Claude Managed Agents
Learn why I moved agent orchestration into durable workflow steps instead of a coordinator agent, and the model, safety-gate, and cost tradeoffs it carried.
- 11 min read
11 min read
Preparing for Mythos: Enterprise CMS Security Checklist
Protect your enterprise CMS stack against the vulnerabilities Mythos exposed with this platform-specific checklist for Sitecore, Umbraco, and Optimizely.
- 10 min read
10 min read
The Agent Operating System: Multi-Agent Pipelines with Claude
Map OS abstractions to Claude Managed Agents architecture and build a three-agent auto-PR pipeline that plans, codes, and reviews autonomously.
- 8 min read
8 min read
Claude Managed Agents: The End of DIY Agent Infrastructure
Discover how Claude Managed Agents replaces months of custom agent infrastructure with a decoupled architecture that cuts time-to-first-token 60%.
- 11 min read
11 min read
WebMCP: How Chrome Turns Websites Into AI Agent APIs
Explore Chrome's WebMCP protocol that lets websites expose structured tools to AI agents, replacing brittle scraping with stable, typed APIs.
- 9 min read
9 min read
Building Plugin Architect: A Zero-Code Claude Code Plugin
Discover how 30KB of prompt engineering replaced a SQLite-backed codebase to become a complete plugin design system for Claude Code's 6 extension points.
- 8 min read
8 min read
Building Plugin GTM: A Go-To-Market Engine Inside Claude Code
Learn how I built a 29-tool MCP server that handles product analysis, GTM strategy, content generation, and launch tracking without leaving the terminal.
