AgentsAtlas is a minimal, opinionated end-to-end workflow for Claude Code (also available as Codex skills). Eight slash commands take a project from setup through planning, execution, review, and milestone completion: /atlas:init, /atlas:plan, /atlas:execute, /atlas:status, /atlas:sync, /atlas:triage, /atlas:review, /atlas:complete. /atlas:execute Runs each 3–5-task plan in a fresh subagent (clean ~200k-token context) so quality doesn't degrade across a long run, while a small CLAUDE.md plus .planning/STATE.md and ROADMAP.md keep session continuity without context bloat. It also ships 16 development "discipline" skills: debugging, testing, verifying, refactoring, security-audit, scaffolding, committing, creating-pr, and more enforcing root-cause-first investigation, evidence-before-claims verification, and test-first development. # Install globally (available in all projects) npx agents-atlas --global # Or install locally (current project only) npx agents-atlas --local MIT licensed; no telemetry, no network calls beyond Anthropic, and no bypass-permissions requirement.