Open Brain
GitHubGitHub RepoNeeds Review

anthropics/claude-quickstarts

Claude Quickstarts offers runnable API examples, including a coding agent that carries progress across sessions with git and an incremental feature list. Builders can inspect concrete patterns before adapting them.

GitHub
Open Source Open MarkdownOpen JSON
Source Summary

The repository collects deployable Claude API examples, including a **two-agent coding workflow** with an initializer and coding agent. It persists progress through git and advances through a feature list across sessions. Other examples cover browser and desktop control, scheduled agents, and managed sessions. Builders running coding agents can study how the coding example divides setup from implementation and resumes work using repository state. The Sentry autofix example adds a failing test, opens a branch for human review, and runs with a spend cap and restricted network access. These are reference implementations with separate setup requirements, and the minimal browser and computer toolset examples are explicitly not production code. The material does not establish how well the examples perform in other environments.

Practical Implication

No separate implication has been recorded yet.

Agent-Ready Context
The repository collects deployable Claude API examples, including a **two-agent coding workflow** with an initializer and coding agent. It persists progress through git and advances through a feature list across sessions. Other examples cover browser and desktop control, scheduled agents, and managed sessions.
Builders running coding agents can study how the coding example divides setup from implementation and resumes work using repository state. The Sentry autofix example adds a failing test, opens a branch for human review, and runs with a spend cap and restricted network access.
These are reference implementations with separate setup requirements, and the minimal browser and computer toolset examples are explicitly not production code. The material does not establish how well the examples perform in other environments.
Connected Context · Feed7 Judgment

The repository offers deployable examples for coding agents, including a two-agent workflow that resumes from git-backed progress, plus a Sentry autofix example with a spend cap and restricted network access. These are implementation references, not evidence of performance or portability.

Context Map
agentcoding#coding-agents#subagents#harness-engineering