jp-executor
Implement a Juspay payment integration in the codebase from a jp-prd PRD and a jp-architecture design + task-checklist. Use after jp-prd and jp-architecture, when the user wants the integration actually built — credentials, SDK install, session/webhook/reconciliation code, DB schema, native setup, portal config, and live tests.
How do I install this agent skill?
npx skills add https://github.com/juspay/merchant-skills --skill jp-executorIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The jp-executor skill is a development tool for implementing Juspay payment integrations. It follows robust security practices by isolating secrets in memory and environment files, and it mandates grounding all code generation in official documentation. The primary security consideration is the surface for indirect prompt injection, as the skill's execution flow is driven by planning artifacts like the PRD and architecture design.
- Socketwarn
1 alert: gptAnomaly
- Snykwarn
Risk: MEDIUM · 2 issues
What does this agent skill do?
JP Executor
You implement a Juspay payment integration directly in the codebase, driven by the upstream planning artifacts in {doc_workspace}: the jp-prd PRD, the jp-architecture design, and its task-checklist.md. You walk the checklist in dependency order, grounding every code/field/method name in re-fetched docs (never memory), and write each task's status back as you go.
Conventions
- Bare paths resolve from skill root;
{skill-root}is this skill's install dir;{project-root}is the project working dir. - Doc-grounding. At implementation time, re-fetch the authoritative Juspay doc pages (recorded in the architecture doc) via
docs-mcpfor exact method/class/field names — never code from memory. - SDK/headless hard gate. No client payment-method code is written until that method's
processpayload page is fetched in-session and its exact request shape is in hand. - Secrets boundary. Credentials (API keys, webhook auth) live only in memory and
.env/secret stores — never in the PRD, architecture doc, logs, or command output. - Debuggability default. Provider/API failures preserve the full provider error body by default for logs/internal handling (subject to secret/PAN redaction), not a collapsed generic message.
- Workspace.
{doc_workspace}is{project-root}/docs/juspay/— readsprd.md,architecture.md, andtask-checklist.mdfrom the upstream skills; writes integration code into the user's codebase, updates taskstatusintask-checklist.mdas it goes, and writes a summary back to{doc_workspace}. In split-repo runs it also reads an incominghandoff-<this_side>.md(if present) and writes the portablehandoff-<other_side>.md. - No config of our own. No settings file, no language/name resolution; reads only the user's repo and the upstream artifacts.
- Production enforced. Always target the production environment; don't ask which environment to use or offer sandbox. Only use a non-production environment if the user explicitly and unpromptedly requests it. Keep key type/stage aligned with the configured host (see
steps/step-02-credentials.md). - Split-repo aware. Build only
side ∈ {this_side, shared}tasks;other_sidetasks are not implemented here — they're carried intohandoff-<other_side>.mdfor the other repo's agent. If an incominghandoff-<this_side>.mdexists, build this side to its contract and verify conformance. Seereferences/split-integration.md. - Stage tracking. This skill owns the
backendandfrontendfunnel phases, keyed to what this run actually builds —startedinsteps/step-01-init.md, terminal insteps/step-09-summary.md, viajuspay_track_integration_stage(fire-and-forget, never blocking). Seereferences/integration-stage-tracking.md.
On Activation
Briefly orient the user: this skill implements a Juspay integration from the jp-prd PRD and jp-architecture design. There is no settings file. Then load ./steps/step-01-init.md, which gates on those inputs.
Execution
Read fully and follow: ./steps/step-01-init.md. The step chain:
step-01-init.md— hard gate (architecture + task-checklist required), parse checklist → queue, codebase re-scan, juspay-mcp mode pickup. (always)step-02-credentials.md— credentials into.env(provision via juspay-mcp or manual dashboard). (adapts to what's needed)step-03-core.md— re-fetch docs, validation layer (if constrained fields), SDK install, core integration + order-status reconciliation. (always)step-04-webhook.md— webhook handler (signature + idempotency). (only if awebhooktask exists)step-05-data-model.md— order/payment schema. (only if adbtask exists)step-06-native-setup.md— mobile native SDK setup. (only on native surfaces)step-07-portal-config.md— dashboard configuration (webhook/return URL) with nav paths + deep links. (only ifmanual-dashboardtasks exist)step-08-test.md— minimal liveness smoke, then hand off tojp-validatefor thorough, stack-aware testing. (only iftesttasks exist)step-09-summary.md— persistent integration summary; finalize checklist statuses. (always)
Thorough testing lives in the dedicated
jp-validateskill (detects the repo's test stack and replicates it, risk-prioritized payment coverage, writestest-report.md). This executor does a liveness smoke and points to it; runjp-validateafter the build.
Steps are conditional. Each non-universal step opens with an APPLICABILITY gate and self-skips when the task-checklist.md (authoritative) and architecture.md contain no work of its kind — no webhook task → no webhook handler; web/API surface → no native setup; nothing to set in the dashboard → no portal step. The executor does only what this integration needs, never manufacturing webhook/DB/native/portal work an integration doesn't require. In a split run, "has an X task" means an X task for this side (side ∈ {this_side, shared}); other_side tasks are already marked skipped in step 1 and never trigger a step here.
This is action-oriented, not a facilitation: implement, but checkpoint before risky/external actions (credential provisioning, DB schema changes, portal configuration, starting the server/tests).
juspay-mcpis auth-guarded; not every developer has dashboard access. The access flow (ask access → log in or manual Q&A) and provenance/reuse rules are inreferences/juspay-mcp.md. Everyjuspay-mcp-derived datum has a manual-input fallback so the flow never blocks.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/juspay/merchant-skills/jp-executor">View jp-executor on skillZs</a>