An Orbit template for turning ambiguous requests into evidence-backed research, architecture decisions, and implementation plans.
Use this template when you want an agent to investigate a feature, refactor, API, framework, or architecture question before implementation begins. The orbit keeps the worker focused on triage, evidence gathering, option comparison, decision records, and phased planning instead of drifting into code changes too early.
Create or enter a Harness runtime, then install the template from GitHub:
harness init
harness install https://github.com/sprneft0604-code/research-plan-orbit.git
orbit enter research-planCheck the installed orbit files:
orbit files research-plan
orbit statusRun the minimum record probe after writing a research-plan record:
bash tools/check-research-plan.sh- A
research-planorbit for one focused research-to-plan pass. - Process notes for triage, research, architecture comparison, planning, and blocked handoff.
- Reusable templates for research-plan records and ADR-style decisions.
- A lightweight probe that checks whether records exist and contain the minimum expected sections.
This template was shaped by comparing several strong research, planning, verification, and spec-first agent workflows. GitHub stars were checked on 2026-04-12.
| Repository | Stars | What it contributed |
|---|---|---|
| superpowers | 147,851 | A strict TDD core: valid RED before production code, minimal GREEN, then refactor. This template keeps that core while using a calmer handoff style for blocked cases. |
| everything-claude-code | 152,062 | A complete TDD agent/workflow shape with evidence and test-layer awareness. This template borrows the evidence mindset without forcing a universal coverage number. |
| spec-kit | 87,242 | A useful model for test-first work driven by project policy and spec clarity. This template keeps room for project-specific testing discipline. |
| get-shit-done | 51,071 | Strong verification and handoff habits. This template adopts explicit blocked/gap recording so failed or uncertain TDD cycles remain useful. |
| gstack | 70,314 | Regression-test and coverage-audit instincts. This template reflects that by making bug fixes prove the regression path. |
| BMAD-METHOD | 44,347 | Broader QA, ATDD, and traceability framing. This template treats those as escalation paths rather than requirements for every small task. |
| OpenSpec | 39,231 | Spec-first discipline for unclear behavior. This template tells agents to stop and clarify when a valid test target cannot be established. |
| oh-my-claudecode | 27,919 | A compact TDD entry-mode idea. This template keeps the installed worker entry short so the agent can get to the test-first loop quickly. |
The result is intentionally focused: deep enough to make implementation safer, but lightweight enough to use before real coding work without turning every task into a long architecture program.