A power-user workflow where the developer treats Claude Code as high-throughput automation infrastructure rather than a conversational assistant. The developer issues terse, goal-oriented commands and expects end-to-end execution across file manipulation, git operations, and codebase exploration. ## Key Characteristics - **Terse, directive style**: "implement this plan and commit", "add this to the config, remove the old one, commit" - **Delegation-heavy tool usage**: Bash (117K invocations), Read (83K), Edit (54K) — Claude as a tireless workhorse - **Multi-file changes as the success pattern**: many successful multi-file edits in a single flow - **Git operations as core workflow**: dozens of git-focused sessions, thousands of commits - **Quick to intervene on overreach**: decisive interruption when Claude misreads scope - **TodoWrite as project management**: Claude as both executor and project manager ## The Trust Boundary The pattern reveals a clear delegation model: freely delegate execution of well-scoped tasks, but retain control over destructive/privileged operations (sudo), scope interpretation (what "implement" means), and quality gates (pre-commit hooks, tests). ## Related Concepts - [[Scope Misinterpretation as Trust Boundary]] — the failure mode of this pattern - [[Self-Healing Git Workflow]] — automating the most common delegation target - [[Multi-Agent Orchestrator as One-Person Dev Team]] — scaling the pattern across agents --- *Source: the author's own Claude Code usage telemetry, published in "Governing Delegation: Using Architecture to Stop AI Scope Creep" (link pending publication)*