EthosPowerAI
Read what GSD is
GSD, short for 'Get Shit Done', is a set of slash commands, agent definitions and file templates that sits inside an AI coding agent and makes it work in planned phases rather than in one long chat. Lex Christopherson, who publishes as TÂCHES and describes himself in the README as a solo developer, created the original repository on 14 December 2025 under the MIT licence. On 20 October 2026 it had 64,269 stars and 5,430 forks; its last commit was dated 31 May 2026, the repository is archived, and its npm package, get-shit-done-cc, stopped at version 1.42.3 on 17 May after 223 published versions.
Follow the move to GSD Core
The archived README now carries one message, that GSD 'continues as GSD Core'. GSD Core, renamed 'Git. Ship. Done.', lives under the Open GSD organisation, with its own site at opengsd.net, also MIT, created on 22 May 2026. On 20 October it had 10,685 stars, 763 forks and 292 open issues, a push the day before and release v1.18.0 from the same day, with releases arriving roughly weekly. The archived repository does not say whether the fork was a handover, and a search on 20 October found nothing published that settles it; the README's single sentence is the whole record. It installs with npx @opengsd/gsd-core@latest on Node 22 or later and supports Claude Code, OpenCode, Codex, Copilot, Cursor, Windsurf and several other agents, with no API key beyond the agent's own.
Trace the loop and its files
A project starts with /gsd-new-project, which asks questions, sends four researcher agents out in parallel, has a synthesiser merge their findings and a roadmapper split the work into phases. It writes PROJECT.md, REQUIREMENTS.md, ROADMAP.md and STATE.md into a .planning folder in the repository. Each phase then runs four commands. /gsd-discuss-phase puts the gray-area decisions to a person and records the answers in CONTEXT.md; /gsd-plan-phase runs a researcher, a planner and a plan-checker in a loop until the plans pass; /gsd-execute-phase hands each plan to an executor, in parallel waves where the plans allow it, and 'each task gets its own atomic commit'; /gsd-verify-work runs a verifier, walks a person through acceptance testing and writes fix plans for anything that fails. /gsd-ship opens the pull request.
Weigh the context-rot claim
The project's central claim is that it 'solves context rot' by running research, planning and execution in fresh-context subagents, so that 'each executor starts with a clean 200k-token context' while the main session stays lean. The mechanism is plausible and the files are real. Neither repository publishes a benchmark, though, so the fair description is a system designed to limit context rot, not one shown to cure it. The README is also candid about cost: the research, plan-check and verifier agents 'add tokens and time', and it offers quality, balanced and budget model profiles instead of a figure.
Decline the skip-permissions advice
The README recommends running Claude Code with --dangerously-skip-permissions so the loop runs without interruption. For a firm's repository that is the wrong trade. An agent that can run any command without asking can also delete, push or fetch anything its shell can reach, and an atomic commit is a poor consolation for a credential sent somewhere it should not go.
Build a commissioning tool with it
The solution applies the loop to a small battery-storage integrator building an internal commissioning-report tool: read the test CSVs, check them against the firm's acceptance limits, draft the report. The engineer answers discuss-phase from the limits and the report template the firm keeps in the repository, and walks through acceptance testing in verify-work before anything ships. Around GSD the design keeps the agent in a sandboxed container with its permission prompts on, pins the GSD Core version, runs continuous integration on the shipped pull request before merge, and starts on the balanced profile with the cost of each phase watched. See the design, its diagram and the demo at the solution page.
The loop is sound. The permissions advice is not.