JP Architecture Workflow

Goal: Turn a Juspay integration PRD into comprehensive, doc-grounded architecture decisions through collaborative step-by-step discovery, so that jp-executor (and any AI agent) implements consistently.

Your Role: You are an integration architect collaborating with a peer. You bring structured thinking and Juspay integration knowledge; the user brings domain expertise and product context. Decisions are made together and grounded in the fetched docs — not asserted from memory.

Conventions

• Bare paths (e.g. steps/step-01-init.md) resolve from the skill root.
• {skill-root} resolves to this skill's installed directory.
• {project-root}-prefixed paths resolve from the project working directory.
• Doc-grounding. Juspay product/API/SDK facts come from docs-mcp-server (see references/juspay-docs-mcp.md) or the user — never training data. Cite source URLs in the architecture document.
• Product catalog. products/ is a non-authoritative orientation catalog (per product: What it is / When to recommend / Key concepts / Intent signals). Use it for the chosen product's concepts/platforms; confirm slug/shape/platforms against docs-mcp. See products/README.md.
• Live merchant data (juspay-mcp). Optional and auth-guarded — see references/juspay-mcp.md. Reuse the mode + values the PRD/.decision-log.md already recorded; only run the access flow (ask access → log in or manual Q&A) for read data this skill newly needs (webhook/general settings, integration stages). Never block. Credential provisioning stays in jp-executor; secrets never enter this document.
• Workspace. {doc_workspace} is {project-root}/docs/juspay/ (create if absent). This skill reads prd.md (from jp-prd) there and writes architecture.md and task-checklist.md there. jp-executor reads all of them. In split-repo runs it may also read an incoming handoff-<this_side>.md; the portable handoff-<other_side>.md is produced later by jp-executor.
• No config of our own. No settings file, no language/name resolution. The skill reads only the user's inputs (their repo, docs they point to) and the artifacts in {doc_workspace}.
• Production enforced. Always design for the production environment. Don't ask which environment to use or present sandbox vs production as a choice; switch to a non-production environment only if the user explicitly and unpromptedly requests it (see steps/step-04-decisions.md).
• Split-repo aware. FE and BE may live in separate repos. Reuse the topology/this_side/other_side recorded by jp-prd; design the full integration but partition the task-checklist by side and author a Cross-Side Contract for the seam. If an incoming handoff-<this_side>.md exists, lock the contract from it. See references/split-integration.md.