Continue the conversation — chat opens pre-seeded with the current signal, caps, and movement.
GSD-Core is the canonical community continuation of GSD (Get Shit Done), the meta-workflow framework that layers structured context engineering and spec-driven development over AI coding runtimes. After solo founder Lex Christopherson (glittercowboy/TÂCHES) executed the $GSD token rug pull (May 22, 2026, ~$500K extracted), deleted all socials, and abandoned the original repo (now archived read-only at 64.6K stars), volunteers under the open-gsd GitHub org migrated the full history — all commits, 394 branches, 229 tags, MIT license, contributor attribution — and stripped the token branding.
IDENTITY (resolved 2026-07-02): open-gsd/gsd-core on GitHub and @opengsd/gsd-core on npm are THE canonical line. 'get-shit-done-redux' was the fork's launch name, renamed ~2026-05-30 — the old repo URL 301-redirects and the old npm package is deprecated with a pointer to @opengsd/gsd-core. There are not two competing open-gsd repos. (Unrelated: npm 'gsd-cc' by pbriese / 0ui-labs is a different project; jnuyens/gsd-plugin is a minor 71-star reimplementation.)
Target user: solo developers and small teams wanting a structured workflow scaffold (planning artifacts, phase orchestration, fresh subagent contexts) over Claude Code or another AI coding runtime. Historically cited among the three leading Claude Code frameworks (with Superpowers and gstack). Explicitly NOT positioned for enterprise.
Within the Workflow Tools cohort it is a different animal from LangGraph: a CLI meta-prompting scaffold for individuals, not a programmable agent-graph runtime with an enterprise deployment story.
Adoption & Proof Points
- **5,761 GitHub stars / 360 forks** (open-gsd/gsd-core, 2026-07-02 via GitHub API) — up from ~940 at the 2026-06-28 gsd.json snapshot (~6x; partly the redux→gsd-core name consolidation)
- **npm @opengsd/gsd-core: 9,894 weekly / 41,292 monthly downloads — now EXCEEDING the abandoned original** (get-shit-done-cc: 5,951 weekly); the interim redux name still draws ~1,077/week from users mid-migration
- **~25 releases in 5 weeks** (v1.2.0 2026-05-31 → v1.6.1 2026-07-01, RC-driven); v1.7.0-rc.1 in flight (CLI version-skew detection, portability AST rules)
- **100-276 commits/week sustained** over the 8 weeks since fork (GitHub participation stats)
- **3 npm publish-authorized maintainers** (trek-e, solvely-colin, jeremymcs); ~12+ non-trivial contributors — but trek-e holds 2,342 commits (dominant single maintainer)
- **openSUSE News coverage (2026-06-26)** treating the fork as a case study in post-rug-pull trust reconstruction
- **No named enterprise customers, no revenue, no funding** — volunteer OSS
- Original gsd-build/get-shit-done: ARCHIVED read-only, 64,638 stars (2026-07-02)
Recommended Use Cases
- Solo developers already running Claude Code (or another supported runtime) who want a structured meta-workflow scaffold — phased orchestration, fresh 200K-token subagent contexts, and `.planning/` artifacts — without building their own context-management discipline from scratch.
- Small teams doing spec-driven greenfield projects where per-phase state files and atomic PR branches (`/gsd-ship`) reduce coordination overhead across async contributors.
- Developers who want to trial a structured AI coding workflow at zero cost and low lock-in: MIT, self-hosted, CLI-only, no web UI or vendor account required.
- Teams evaluating post-rug-pull OSS continuations as a supply-chain governance case study — the open-gsd migration (full commit/branch/tag history, contributor attribution preserved) is documented and publicly auditable.
- **Not recommended** for teams with enterprise security requirements: advisory-only runtime hooks, no RBAC/SSO/audit, bus-factor-1 volunteer governance, and open npm supply-chain exposure to the rug-puller's retained publish keys make this unsuitable for production environments with compliance obligations.
- **Not recommended** as a stable dependency in automated CI pipelines until runtime hook enforcement moves from Phase-3 roadmap to shipped — a prompt-injection event currently yields full local shell execution with no circuit breaker.
Risks & Limitations
- **Advisory-only runtime security model (OPEN):** agents hold full shell access; the four injection/out-of-scope hooks warn but never block — a successful prompt injection yields full local code execution with no circuit breaker. CI-side gates hardened; runtime hooks did not (fork's own baseline doc lists it as future Phase 3).
- **Supply-chain exposure (OPEN):** the rug-puller retains publish rights on live npm packages (get-shit-done-cc installable; @gsd-build/sdk not deprecated). A malicious publish to either would hit users who never migrated.
- **Bus factor ~1:** trek-e dominates commits; no GOVERNANCE.md, MAINTAINERS.md, or CODE_OF_CONDUCT.md; volunteer team, not a foundation.
- **No commercial backstop:** unfunded, no revenue model; the original line's 'funding' was the rug-pulled meme coin (funding-concerns cap).
- **Provenance shadow:** six weeks of post-fraud history; community trust is being rebuilt from a founder exit-fraud collapse.
- **No enterprise governance:** RBAC/SSO/audit/certs absent; solo-dev / small-team positioning is explicit.
- **Capability counts partially unverified:** 33 subagents / parallel waves / 15 runtimes are carried from the prior GSD eval, not re-verified against current gsd-core source (docs enumerate 9+ runtimes).
- **--dangerously-skip-permissions** reportedly remains the documented autonomous building-phase workflow (carried from prior eval, not re-verified).
Capabilities & Integration
Inherited and actively maintained meta-workflow design: per-phase orchestrators writing state to disk; each atomic plan executed in a fresh 200K-token subagent session (main session held at 30-40% utilization); specialized subagents with parallel dependency waves; /gsd-map-codebase architecture analysis; .planning/ Markdown/JSON artifacts; /gsd-ship producing clean PR branches with atomic commits handed to external CI; gsd-sdk headless CI/CD surface.
Runtime support: installer translating across AI coding runtimes — docs explicitly enumerate Claude Code, OpenCode, Gemini CLI, Kimi CLI, Kilo, Codex, GitHub Copilot, Cursor, Windsurf 'and more' (the '15 runtimes' figure is carried from the pre-fork eval, not re-verified).
Fork-era additions (from release history): Claude Sonnet 5 support (v1.6.1, 2026-07-01); capability trust gates with a user-owned consent store gating third-party capability activation and fail-closed handling of a corrupt capability ledger (v1.6.0, 2026-06-24 — scoped to capability loading, not core executor hooks); package-legitimacy/slopcheck dependency screening; CLI version-skew detection + portability AST rules (v1.7.0-rc.1). Language mix: JS 82.7% / TS 16.8%.
Distribution: npx @opengsd/gsd-core (npm, scoped). CLI/terminal-only — no web UI, no IDE extension, no REST API.