Institutional data and research workflows

Make rigorous work
with AI, responsibly.

Compound GPID structures AI-assisted development for data scientists, economists, and technical teams working under institutional constraints. It connects planning, implementation, review, and reusable project knowledge while keeping human judgment visible, testable, and accountable.

Get started Read the philosophy

Core workflow Verified steps

01Understand/cg-brainstorm
02Plan/cg-plan
03Deliver/cg-work
04Assure/cg-review
05Resolve/cg-fix-triage
06Remember/cg-compound

Human judgment

Specify intent, challenge plausible output, and keep the ability to explain the work you accept.

Independent evidence

Separate generation from verification with tests, references, cross-checks, and review matched to risk.

Shared reasoning

Capture decisions, known risks, and verified lessons so institutional knowledge does not disappear into chat.

Choose your route

Start with the task,
then disclose the detail.

Using AI responsiblyUnderstand the philosophy and its human boundaries New hereInstall, configure, and run a first workflow Choose a suiteCompare technical and research workflows Ready to workChoose a task-oriented workflow Writing codeFind analytical and technical skills Assessing fitUnderstand governance and limitations Something brokeStart a safe recovery path

What the plugin provides

A structure for responsible AI-assisted work.

The canonical .github/ source is owned through a three-layer registry: kernel, reusable capability packs, and independent technical /cg-* and research /cr-* suites. Release-validated native trees support GitHub Copilot, Claude Code, Codex, OpenCode, and Kilo. Read the Modular Guide to choose a suite.