03 Agent tooling
MAINFRAME
Shell work proposed by a Pi agent needs a clear permission boundary. MAINFRAME provides shell policy, checkpoints and status a person can inspect.
Agent tooling · shell policy and recovery workflows
Open source

Project record
Public-facing client work, open-source tooling and original engineering experiments.
Client entries cover professional context and public websites. Technical walkthroughs belong to my independent work and original demonstrations.
Not sure where to start? Ask Project Advisor . Explore a small first build, its tradeoffs and a testable next step.
03 Agent tooling
Shell work proposed by a Pi agent needs a clear permission boundary. MAINFRAME provides shell policy, checkpoints and status a person can inspect.
Agent tooling · shell policy and recovery workflows
Open source

Earlier portfolio entries are omitted where their scope or outcomes have not been reviewed. Ask about other work ↗
From a question to the implementation
The original notes explain choices behind my independent work. Business decision exercises provide a separate starting point for technical leadership.
Shell policy, checkpoints, readiness checks and the limits of the protected Pi routes.
Picking a model is the smallest decision in agentic coding. A fractional CTO's guide to alignment, vertical slices, token budgets and guardrails, with a checklist for founders and engineering leads.
Server-rendered content, optional integrations and protected publishing boundaries.
How this portfolio separates a local state simulation from a private realtime workflow, with explicit ownership, subscription readiness and recovery after an uncertain command.
Connect a business decision to technical direction, architecture and hands-on software development.
Compare leadership responsibilities, build-versus-buy options and a roadmap before a rewrite.
Engineering with synthetic data
Inspect duplicate requests and failed notifications in a fictional telescope lending desk. The experiment is independent of client and employer systems.
Open the lending labProject advisor / a decision workshop
A useful first version starts with a smaller, sharper question.
Choose a project pattern, make up to three decisions and leave with a scoped proposal: what to test, what to leave out, and what would make you change course.
Or talk to me directly ↗Your selected pattern and choices shape the recommendations. Your description appears in the brief for you to review; the planner does not analyze it. Everything stays in this browser tab, and refreshing clears the session.