By lotea-be
QRSPI — an eight-stage spec-driven workflow (Questions → Research → Design → Structure → Plan → Worktree → Implement → PR) layered on OpenSpec. Ships the /qrspi-* commands, the opsx-* OpenSpec helpers, seven stage subagents, and the supporting skills.
Archive a completed QRSPI change after its PR has merged. Thin wrapper over OpenSpec's archive skill — moves the change folder under archive/ with a date prefix and syncs delta specs into the main specs.
QRSPI stage D. Delegates to the qrspi-designer subagent to produce design.md (~200 lines). HUMAN REVIEW IS REQUIRED before stage S.
QRSPI post-PR fix loop. Delegates to the qrspi-implementer subagent in FIX MODE to resolve a single post-PR follow-up (reviewer open issue or retro code flag) while keeping code, tests, and the change's DELTA spec in sync. Ticks followups.md and commits atomically on the PR branch.
QRSPI stage I. Delegates to the qrspi-implementer subagent to write code one vertical slice at a time, ticking tasks.md as it goes. Stops at each slice checkpoint for human verification.
Bootstrap OpenSpec in this repo if it has not been done yet. Runs `npx @fission-ai/openspec init`. Safe to run more than once — will detect existing setup.
QRSPI stage Q. Turns a vague feature request into a concrete list of technical questions the codebase must answer. Writes openspec/changes/<id>/questions.md.
QRSPI stage R. Read-only investigator that maps the current codebase areas relevant to a change. The change ticket is hidden from you on purpose. Writes openspec/changes/<id>/research.md.
QRSPI stage D. Produces the ~200-line design.md that surfaces the agent's assumptions for human review. This is the "brain surgery" stage. Writes openspec/changes/<id>/design.md. Implementation is BLOCKED until a human approves this artifact.
QRSPI stages S and W. Turns an approved design into a Structure outline (proposal.md + specs/) and a Worktree plan of vertical slices. Owns openspec/changes/<id>/proposal.md, specs/, and worktree.md.
QRSPI stage P. Read-only on code. Turns the Structure + Worktree into a tactical task checklist. Writes openspec/changes/<id>/tasks.md.
Rules for keeping agent context windows lean (target under 40%, reset at 60%) and using subagents as context firewalls rather than personas. Load this when planning a multi-stage workflow, starting a long session, or deciding whether to delegate work to a subagent.
Implement tasks from an OpenSpec change. Use when the user wants to start implementing, continue implementation, or work through tasks.
Archive a completed change in the experimental workflow. Use when the user wants to finalize and archive a change after implementation is complete.
Enter explore mode - a thinking partner for exploring ideas, investigating problems, and clarifying requirements. Use when the user wants to think through something before or during a change.
Propose a new change with all artifacts generated in one step. Use when the user wants to quickly describe what they want to build and get a complete proposal with design, specs, and tasks ready for implementation.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A spec-driven, multi-stage workflow for Claude Code
and GitHub Copilot, built on top of
OpenSpec. QRSPI breaks a feature from vague
request to merged PR into eight reviewable stages, each producing one durable artifact
under openspec/changes/<id>/. A human approval gate sits at the Design stage, so the
expensive thinking is reviewed before any code is written.
This repository is the source of truth for the QRSPI kit. The Claude Code artifacts
under claude/ are authoritative; the Copilot artifacts under
copilot/ are generated from them (see
Two tools). Claude Code installs it as a plugin (from the
lotea-be/ai-agent-marketplace marketplace); GitHub Copilot installs the generated
artifacts into ~/.copilot/ with a script. Either way, the kit is available in every
repository on your machine.
The kit is stack-agnostic. It references a project's "stack-cheatsheet" skill generically and degrades gracefully when a repo doesn't provide one — so the same workflow drives a .NET/Blazor app, a TypeScript service, or anything else.
| # | Stage | Command | Artifact | Notes |
|---|---|---|---|---|
| 1 | Questions | /qrspi:questions <id> | questions.md | Turns a vague request into concrete technical questions. |
| 2 | Research | /qrspi:research <id> | research.md | Read-only map of the current codebase. The ticket is hidden by design. |
| 3 | Design | /qrspi:design <id> | design.md | The "brain surgery" stage. ⛔ HUMAN APPROVAL REQUIRED before stage 4. |
| 4 | Structure | /qrspi:structure <id> | proposal.md + specs/ | Canonical proposal + OpenSpec spec deltas. |
| 5 | Worktree | /qrspi:worktree <id> | worktree.md | Vertical slices, not horizontal layers. |
| 6 | Plan | /qrspi:plan <id> | tasks.md | Canonical numbered task list. |
| 7 | Implement | /qrspi:implement <id> | code + tests | One slice at a time; stops at each checkpoint. |
| 8 | PR | /qrspi:pr <id> | PR description | Read-only review + final checklist. |
Helpers: /qrspi (print the stage map), /qrspi:init (bootstrap openspec/ +
templates — per-repo onboarding), /qrspi:stack (bootstrap this repo's
stack-cheatsheet skill — per-repo onboarding), /qrspi:followup <id> (post-PR fix loop),
/qrspi:archive <id> (archive a change after its PR merges),
/qrspi:retro <id> <stage> (retrospective that improves the prompts themselves).
Each artifact follows a canonical OpenSpec shape — see
openspec-templates/.
qrspi/
claude/ # SOURCE OF TRUTH — Claude Code plugin payload
agents/ # 7 subagent definitions (qrspi-questioner … qrspi-reviewer)
commands/ # /qrspi:* slash commands (+ opsx wrappers)
skills/ # workflow + convention skills (stack-agnostic)
copilot/ # GENERATED from claude/ — GitHub Copilot artifacts (mirror ~/.copilot/)
agents/ # *.agent.md (custom agents)
prompts/ # *.prompt.md (slash prompts)
instructions/ # *.instructions.md (referenced on demand)
openspec-templates/ # the 5 canonical artifact templates (tool-independent, shared)
sync-copilot.ps1 # deterministic claude/ -> copilot/ generator
install.ps1 / install.sh # installs the Copilot kit into ~/.copilot (Claude = plugin)
uninstall.ps1 / uninstall.sh # removes only the files this kit ships
.claude/ # kit-DEV tooling, project scope, NOT shipped to users:
# /qrspi-sync-copilot command + skill (only useful in THIS repo)
README.md
copilot/ mirrors ~/.copilot/, so the Copilot install is a straight copy; claude/
is the Claude Code plugin payload, delivered via the marketplace. Never hand-edit
copilot/ — edit claude/ and run /qrspi-sync-copilot (see
Two tools).
claude/{agents,commands,skills}/ formats are Claude Code's.copilot/{agents,prompts,instructions}/ (.agent.md / .prompt.md /
.instructions.md).npx) — the OpenSpec CLI runs via
npx @fission-ai/[email protected] (pinned). Used to bootstrap openspec/ and to
openspec validate spec deltas. The kit's claude/commands/opsx/ and
claude/skills/openspec-* are generated by this exact version — bump the pin andnpx claudepluginhub lotea-be/ai-agent-marketplace --plugin qrspiHarness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.