juspay

Drive one AI agent from another through kolu's terminals: spawn a Claude Code / Codex / opencode session in a PTY, prompt it, watch the screen for its reply, read it, and prompt again. PRIMARY PATH: the kolu MCP server's tools (`lifecycle_create`/`lifecycle_sendInput` with named keys, the `wait_outputSettled`/`wait_agentState` done-signals, `screen_text`, the `terminals` resource) — configured in this repo's .mcp.json as `kolu mcp`. FALLBACK when the MCP is unavailable: the `kaval-tui`/`padi-tui` CLIs, same verbs, same discipline. Triggers on "drive another agent", "send a prompt to a terminal agent", "have one agent prompt another", "agent drives agent", "orchestrate agents in terminals", "make Claude drive Codex", "prompt the agent running in that terminal", or wiring a loop where one coding agent supervises another.

Running memory of the coordination rules for an orchestrator agent driving implementing agents on kolu terminals — authorization boundaries, the dispatch protocol (MCP-first via the kolu skill, with its TUI fallback), verification discipline, and how to communicate with the human. Load whenever coordinating a multi-agent campaign (dispatching briefs to agents in kolu PTYs, tracking their PRs, verifying their claims), or when acting as the coordinator in a padi/surface-style campaign. Triggers on "orchestrate the agents", "coordinate the campaign", "dispatch to the agents", "act as coordinator", or driving multiple implementing agents from one supervising session.

Running memory of the coordination rules for an orchestrator agent driving implementing agents on kolu terminals — authorization boundaries, the dispatch protocol (MCP-first via the kolu skill, with its TUI fallback), verification discipline, and how to communicate with the human. Load whenever coordinating a multi-agent campaign (dispatching briefs to agents in kolu PTYs, tracking their PRs, verifying their claims), or when acting as the coordinator in a padi/surface-style campaign. Triggers on "orchestrate the agents", "coordinate the campaign", "dispatch to the agents", "act as coordinator", or driving multiple implementing agents from one supervising session.

Modern, interactive alternative to `/do` — clarify intent up front, then take a task end-to-end with a serial AI review gauntlet (lens debate (lowy ⇄ hickey) → codex debate → simplify → code-police, each editing the branch in turn) → CI → evidence. ONLY invoke when the user explicitly types `/be` or `$be`; never auto-select from a natural-language request.

Run an automated codex⇄Claude debate to consensus — no round cap, no deadlock exit. Two explicit subcommands. `review` (also the bare/back-compat default) — codex critiques the current diff and Claude (author) fixes/disputes, looping until they agree, and the trail is committed + posted to the PR. `answer` — Claude and codex each answer a freeform prompt, then cross-check until they agree, and a unified answer is returned. The debate runs over a LIVE codex session in a split terminal beside you (driven per the /kolu skill), not a headless workflow. Use when the user types `/codex-debate`, asks to "have codex review this", "run the codex debate", "review this PR with codex", "argue this with codex until you agree", or passes a question to "have Claude and codex debate/answer until they agree".

Run an automated codex⇄Claude debate to consensus — no round cap, no deadlock exit. Two explicit subcommands. `review` (also the bare/back-compat default) — codex critiques the current diff and Claude (author) fixes/disputes, looping until they agree, and the trail is committed + posted to the PR. `answer` — Claude and codex each answer a freeform prompt, then cross-check until they agree, and a unified answer is returned. The debate runs over a LIVE codex session in a split terminal beside you (driven per the /kolu skill), not a headless workflow. Use when the user types `/codex-debate`, asks to "have codex review this", "run the codex debate", "review this PR with codex", "argue this with codex until you agree", or passes a question to "have Claude and codex debate/answer until they agree".

Run /be's review gauntlet SERIALLY — architecture-first-principles FIRST (framework/structure diffs), then /lens-debate (lowy ⇄ hickey), then /codex-debate, then /simplify, then code-police, each editing and committing on the live branch in turn. Use from /be §4, or when the user asks to "run the review gauntlet". Requires Claude Code's Skill tool.

Run a structural-review debate between two lenses — lowy (volatility-based decomposition) and hickey (structural simplicity) — on the current diff. Each reviews independently, then the reviews are reconciled (cross-lens agreements settle with zero debate) and only genuinely contested findings debate in parallel per-file threads until the lenses agree; the agreed fixes are applied. Use when the user types `/lens-debate`, or asks to "have lowy and hickey review this", "run the lens debate", "debate this diff structurally", or "argue the structure of this PR until the lenses agree".

Use when generating connector coverage or capability metrics for meetings, manager/stakeholder updates, retrospectives, or "what changed" / month-over-month / since-release comparisons in the hyperswitch-prism (UCS) repo. Covers coverage diffs over a time window, current-state snapshots, merged-PR activity summaries, and per-connector / per-flow drilldowns. Triggers: "data for the meeting", "how much did coverage change last month", "what did we ship", "supported vs last release".

Reviews pull requests in the hyperswitch-prism (UCS) Rust codebase using a strict, fail-closed, scenario-aware review system. Classifies PRs into connector, core-flow, proto, server, SDK, CI/security, and GRACE-generated scenarios, then dispatches specialist subagents per scenario. Use when reviewing any PR, batch-reviewing open GRACE PRs, or re-reviewing after author updates.

Embeds prism into any app and enables to connect to any payment processor

Helps developers integrate with the Hyperswitch Prism SDK across different programming languages and payment scenarios. Provides skills for setting up payment clients, processing payments, handling errors, routing between connectors, configuring connectors, and processing refunds. Supports Python, Node.js, Java/Kotlin, and Rust SDKs.

Adds payment method support (Card, Wallet, Bank Transfer, UPI, BNPL, etc.) to an existing connector in the connector-service (UCS) Rust codebase. Modifies the Authorize flow transformers to handle new payment method data types. Use when a connector exists with Authorize flow but needs to support additional payment methods.

Adds one or more payment flows (Authorize, Capture, Refund, Void, PSync, RSync, webhooks, etc.) to an existing connector in the connector-service (UCS) Rust codebase. Use when a connector already exists but is missing specific flow implementations. Handles dependency validation and sequential implementation order.

Generates a technical specification for a payment connector in two phases: (1) discover and verify the connector's official API documentation links, (2) feed those links into the grace techspec CLI to produce a structured spec. Each phase can be delegated to a subagent. Use before implementing a new connector with the new-connector skill.

Implements a new payment connector from scratch in the connector-service (UCS) Rust codebase. Creates connector foundation and implements all 6 core payment flows (Authorize, PSync, Capture, Refund, RSync, Void). Use when integrating a new payment gateway that does not yet exist. Requires a technical specification at grace/rulesbook/codegen/references/{connector_name}/technical_specification.md.

Use this when building a Bun (bun.lock) project with Nix. Covers bun2nix, the bun.nix dependency workflow, the bun --compile top-level-await pitfall, and a CI drift check.

Use this when setting up CI for a GitHub repository — offers GitHub Actions or Vira depending on the project

Use this when a flake's `nix develop` / `direnv allow` / `nix flake archive` is slow on a fresh checkout (the "first time takes 10 minutes" complaint). Diagnoses where the time actually lives and how to shrink the flake.lock input graph without changing build outputs.

Write a programming essay or blog post in the voice of the canon — Spolsky, Yegge, Graham, Mickens, Dijkstra, Brooks, Nystrom, Kleppmann, patio11. Invoke when the user wants to argue an idea about software, architecture, languages, or the craft — not a debugging war story (use debugging-story for that), not a tutorial, not a release note. The audience is working developers worldwide with taste and strong opinions of their own.

Use this when diagnosing or fixing a user's Nix installation — checks flakes, version, caches, max-jobs, direnv, rosetta, trusted-users, and shell config

Conventions for building Leptos CSR apps with Nix (crane + Trunk).

Run cargo-watch in the background for continuous clippy feedback during code editing.

Techniques for creating deterministic terminal demo screencasts with VHS

Use this when writing or reviewing Haskell code. Covers error handling, type safety, idiomatic patterns, HLint compliance, Aeson usage, and testing.

Do a task end-to-end — implement, PR, CI loop, ship. ONLY invoke when the user explicitly types `/do` or `$do`; never auto-select from a natural-language request, even one that sounds like an end-to-end task.

Evaluate code (especially LLM-generated) for structural simplicity using Rich Hickey's "Simple Made Easy" framework. Use this skill whenever reviewing a PR, diff, or code snippet for accidental complexity — particularly when the code was generated by an AI coding assistant and line-by-line review isn't feasible. Also use when the user asks about complecting, simplicity vs. easiness, structural coupling, or concept deduplication. Trigger on phrases like "is this simple", "does this complect", "review for complexity", "structural analysis", or any reference to Hickey, Simple Made Easy, or grey-box review.

Evaluate architecture and module boundaries for volatility-based decomposition using Juval Lowy's framework (from "Righting Software", building on Parnas 1972). Use when reviewing module splits, service boundaries, new abstractions, or any decomposition decision. Trigger on phrases like "where should this boundary be", "how to split this", "module boundaries", "encapsulate change", "volatility", or references to Lowy, Parnas, or "Righting Software". Complements /hickey (interleaved concerns) with a different lens (change encapsulation).

Enter talk mode — conversation and research, no repo changes. ONLY invoke when the user explicitly types `/talk` or `$talk`; never auto-select from a natural-language question or design discussion.

Review code for quality, simplicity, and common mistakes before declaring work complete.

Vira's design system — TailwindCSS conventions, color palette, typography, spacing, components, and accessibility rules. Use when working on UI components, styling, or layout in this project (Lucid HTML in Haskell sources under packages/vira).

Reference for the `odu` runner — how to invoke a full pipeline, a single recipe, or a platform-pinned node, and how to attach to a live run, from a project whose CI odu runs. Trigger when the user asks to "run CI", "run the pipeline", "re-run a check", to run named lanes or recipes (e.g. "run fmt and nix", "just the e2e lane", bare selectors like `fmt`/`nix`/`e2e`), or names a recipe by `<recipe>@<platform>`. This skill — not a repo's local `just ci` / `just <recipe>` — is how an odu-run request is served.

Run an automated codex⇄Claude debate to consensus — no round cap, no deadlock exit. Two explicit subcommands. `review` (also the bare/back-compat default) — codex critiques the current diff and Claude (author) fixes/disputes, looping until they agree, and the trail is committed + posted to the PR. `answer` — Claude and codex each answer a freeform prompt, then cross-check until they agree, and a unified answer is returned. The debate runs over a LIVE codex session in a split terminal beside you (driven per the /kolu skill), not a headless workflow. Use when the user types `/codex-debate`, asks to "have codex review this", "run the codex debate", "review this PR with codex", "argue this with codex until you agree", or passes a question to "have Claude and codex debate/answer until they agree".

Running memory of the coordination rules for an orchestrator agent driving implementing agents on kolu terminals — authorization boundaries, the dispatch protocol (MCP-first via the kolu skill, with its TUI fallback), verification discipline, and how to communicate with the human. Load whenever coordinating a multi-agent campaign (dispatching briefs to agents in kolu PTYs, tracking their PRs, verifying their claims), or when acting as the coordinator in a padi/surface-style campaign. Triggers on "orchestrate the agents", "coordinate the campaign", "dispatch to the agents", "act as coordinator", or driving multiple implementing agents from one supervising session.

Provision and drive a `pu` box — an Incus container used as a clean Linux host for CI, builds, and evidence capture. Use when you need to run something on a fresh remote box instead of the user's machine: `nix run` a build, run CI against a real host, capture screenshots/video off-machine, or reproduce on a pristine environment. Covers create/connect/scp/destroy, running remote commands, copying artifacts back, and the no-egress failure mode. Triggers on "pu box", "spin up a box", "run this on a box", "ephemeral host", "pu create/connect/destroy".

odu MCP server launcher — drive CI from a coding agent. `bin/serve` resolves odu via Nix and runs `odu mcp` in the cwd. See the repo README for the tools/resources and override knobs.

Reference for the `odu` runner — how to invoke a full pipeline, a single recipe, or a platform-pinned node, and how to attach to a live run, from a project whose CI odu runs. Trigger when the user asks to "run CI", "run the pipeline", "re-run a check", to run named lanes or recipes (e.g. "run fmt and nix", "just the e2e lane", bare selectors like `fmt`/`nix`/`e2e`), or names a recipe by `<recipe>@<platform>`. This skill — not a repo's local `just ci` / `just <recipe>` — is how an odu-run request is served.

Create, update, or validate a Juspay payment-integration PRD. Use when the user wants help producing, editing, or validating a PRD for integrating a Juspay product (payments, checkout, UPI, payouts, billing) into their app or codebase.

Design the architecture for a Juspay payment integration — collaborative, step-by-step decisions that ground the integration in real Juspay docs and produce a consistent implementation guide. Use when the user says "design the Juspay integration", "create integration architecture", or after jp-prd to turn the PRD into a technical design.

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.

Test/validate a Juspay payment integration after jp-executor has built it. Detects the repo's existing test stack and replicates it (Playwright, Cypress, Jest, Vitest, pytest, supertest, …), falling back to inline curl/bash when none exists. Risk-prioritized coverage of order/session creation, server-to-server status reconciliation, webhook signature + idempotency, per-method process payloads, constraint boundaries and error-code paths — backend and frontend/SDK as the built surface dictates. Ends with a traceability matrix, a PASS/CONCERNS/FAIL quality gate, and a written test report. Use after jp-executor, or standalone against an already-built integration.

Entry point for Playwright test automation. ALWAYS delegates to orchestrator.md. The orchestrator (YOU) detects execution mode (full pipeline or heal-only) and manages the complete workflow including setup, execution, summary, bug reports, and cleanup. Triggers on phrases like "generate tests", "create tests", "run tests", "test flow", "end-to-end test", "e2e test", "test PR", "test module", "test scenario", "analyze for testing", "generate test cases", "write test code", "create test file", "heal tests", "fix failing tests", "debug tests", "repair tests".

Review a GitHub PR by fetching its diff, analyzing it with momus (expert reviewer), and producing structured inline comments with file paths, line numbers, and severity levels. MUST USE when the user shares a GitHub PR link and asks for a review, says "review this PR", "look at this pull request", "give me feedback on PR

End-to-end PR workflow for the hyperswitch-control-center repo. Use this skill whenever the user wants to raise a PR, submit changes, open a pull request, commit and push, or ship their work. Triggers on phrases like "raise a PR", "open a PR", "submit my changes", "commit and push", "ship this", "create a pull request". Even if the user just says "PR please" or "push my changes", use this skill.

Use this when setting up CI for a GitHub repository — offers GitHub Actions or Vira depending on the project

Use this when writing or editing a flake.nix. Covers flake-parts, perSystem, formatter, shell scripts, and package conventions.

Reference for the `justci` runner — how to invoke a full pipeline, a single recipe, or a platform-pinned node from a project that depends on `juspay/justci`. Trigger when the user asks to "run justci", "run the pipeline", "re-run a check", or names a specific recipe by `<recipe>@<platform>`.

Bootstrap or update srid/agency in this project — run apm via uvx, configure apm.yml, install skills, draft workflow instructions. Use for first-time setup or to refresh an existing install.

Local codebase analysis for docs

Wire a self-hosted Juspay Decision Engine into a merchant's own payment orchestrator so PSP/gateway selection is delegated to Decision Engine. Use this skill whenever the user is working in their orchestration / payment-routing / checkout backend and wants to call Decision Engine's `decide-gateway` (to pick the best PSP before authorizing) and `update-gateway-score` (to feed the auth/charge outcome back), or mentions "Decision Engine", "decide gateway", "gateway score", "PSP routing", "dynamic routing", "success-rate routing", "which gateway should I use", or hooking a Vault/tokenized-card checkout up to smart routing. Trigger this even when the user only describes the goal ("route this payment to the best acquirer", "add smart routing before we charge the PSP") without naming the APIs. This skill covers discovering the authorize/charge path in the orchestrator, inserting the two calls in the right place, mapping the tokenized-card + BIN metadata into the request, failing open safely, and certifying the integ

Guide for using the NeuroLink SDK and CLI. Invoke when users ask how to use neurolink, integrate AI providers, add MCP tools, configure RAG, set up memory, deploy servers, or work with multimodal content. Covers SDK, CLI, providers, tools, and enterprise features.