Complete procedures with a way back.

Turn a proven workflow into a narrowly triggered SKILL.md, package it in a versioned Codex plugin, add fail-closed policy hooks and least-privilege MCP dependencies, test activation and denial paths, distribute through a controlled marketplace, and revoke safely.

Build and verify a default-off Claude Code plugin that packages a manually invoked review skill, a genuinely read-only plugin subagent, a local stdio MCP policy tool, exact runtime dependencies, and a distributable marketplace entry without embedding credentials or pretending that descriptive annotations enforce authorization.

Build a ChatGPT-compatible MCP service with read and write tools, an optional sandboxed UI component, OAuth-aware authorization, privacy disclosures, host testing, and an evidence-backed submission checklist.

Evaluate an MCP server as executable supply chain and delegated authority, pin its identity, constrain a local runtime or verify a remote trust endpoint, configure a positive tool allowlist, review OAuth scopes, observe real effects, test prompt injection, set per-tool approvals and rehearse complete removal.

Design and deploy a remote TypeScript Model Context Protocol server whose tools, resources and prompts have closed schemas, per-handler authorization, OAuth protected-resource discovery, exact audience and scope validation, isolated downstream credentials, bounded transport, audit evidence and a rehearsed rollback.