catalyst-basics — Claude Code skill: Catalyst project setup
A skill that walks the agent through Zoho Catalyst project setup: directory layout, environments, CLI commands and every project ID. Handy when you need a Table ID or a look at .catalystrc.
npx skills add catalystbyzoho/agent-skills --skill catalyst-basics -a claude-codeWhat catalyst-basics does
This skill covers the scaffolding side of Catalyst work. Ask about .catalystrc, catalyst.json, the gap between Development and Production, or where a given ID lives, and the agent pulls the matching reference file: quick start, console navigation, project structure, CLI reference, service selection, plus setup notes for Claude, Cursor and GitHub Copilot.
MCP comes first. The agent looks for CatalystbyZoho_* tools and, when they exist, reads org ID, project ID, table names and ZAIDs straight from them instead of asking you to copy values out of the console. Before catalyst init it runs catalyst whoami and compares the org against CatalystbyZoho_List_All_Organizations, since MCP and the CLI can sit on different accounts; a mismatch means catalyst login --force first.
The skill refuses to hand-write .catalystrc and catalyst.json — those come from catalyst init --org <orgId> -p <projectId> -ni, never from an interactive run. Without Zoho MCP connected, set that up before anything else, or the agent falls back to reading local project files.
What it helps with
- locate a Project ID, ZAID or Table ID without opening the console
- understand what
.catalystrcandcatalyst.jsonhold - tell Development and Production environments apart
- pick the right Catalyst service for a pattern
- wire up Zoho MCP in Claude, Cursor or GitHub Copilot
Example prompts for the agent
- Help me start a Catalyst project — which files should show up in the directory?
- Where do I get the Table ID for my table, and can MCP fetch it?
- Explain how the Development environment differs from Production in Catalyst
How to install catalyst-basics
1Skills CLI (Vercel)
npx skills add catalystbyzoho/agent-skills --skill catalyst-basics -a claude-code2Manually: put the files in place
~/.claude/skills/catalyst-basics/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/catalyst-basics/
1Skills CLI (Vercel)
npx skills add catalystbyzoho/agent-skills --skill catalyst-basics -a codex2Manually: put the files in place
.agents/skills/catalyst-basics/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add catalystbyzoho/agent-skills --skill catalyst-basics -a cursor2Manually: put the files in place
.cursor/skills/catalyst-basics/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install catalystbyzoho/agent-skills catalyst-basics2Manually: put the files in place
.github/skills/catalyst-basics/SKILL.md1Skills CLI (Vercel)
npx skills add catalystbyzoho/agent-skills --skill catalyst-basics -a gemini-cli2Platform CLI
gemini skills install https://github.com/catalystbyzoho/agent-skills --path skills/catalyst-basicsManually: put the files in place
.devin/skills/catalyst-basics/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add catalystbyzoho/agent-skills --skill catalyst-basics -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about catalyst-basics
How do I install catalyst-basics in Claude Code?
Run npx skills add catalystbyzoho/agent-skills --skill catalyst-basics -a claude-code. The skill file then lands at ~/.claude/skills/catalyst-basics/SKILL.md.
What does catalyst-basics do?
It guides the agent through Catalyst project setup: directory structure, environments, CLI commands and resource IDs. Depending on the question it loads a reference file, from quick start to service selection.
Does the skill need MCP?
It checks for CatalystbyZoho_* tools first and uses them to fetch org ID, project ID and other identifiers. With no Zoho MCP connection it reads .catalystrc and catalyst.json from the project directory instead.
Source text
SKILL.md · Apache-2.0
--- name: catalyst-basics description: "Core Catalyst project setup — directory structure, environments, CLI commands, and all Catalyst IDs (Project ID, ZAID, Table ID, Segment ID, Org ID). Trigger on 'start a Catalyst project', 'what is .catalystrc', 'where do I find my Table ID', 'difference between Development and Production', or any Catalyst ID or CLI question." metadata: version: "2.0.2" --- > **⚠️ PRE-FLIGHT CHECK:** Before creating any project files, confirm that `.catalystrc` and > `catalyst.json` exist in the project directory. If they don't: > 1. Use MCP (`CatalystbyZoho_List_All_Organizations` → `CatalystbyZoho_List_All_Projects`) to get the org ID and project ID. > 2. Run: `catalyst init --org <orgId> -p <projectId> -ni` > 3. **Never ask the user to run `catalyst init` interactively. Never create these files manually** — they are CLI-generated and must come from the CLI. ## How It Works 1. **MCP check first** — Before reading any local files, look for `CatalystbyZoho_*` tools. If available, use MCP to get org ID, project ID, and all resource IDs instead of asking the user. 2. **Account alignment check — do this before `catalyst init`** — MCP and CLI can be authenticated as different accounts. Before running `catalyst init --org <id>`, verify they match: ```bash catalyst whoami ``` Compare the org name shown to the result of `CatalystbyZoho_List_All_Organizations`. If they differ, run `catalyst login --force` to re-authenticate the CLI before proceeding. 3. **New to Catalyst?** — If the user is setting up Catalyst for the first time or asking "how do I start", load `references/quick-start.md` for the full walkthrough. 4. **Console navigation** — If the user asks how to find IDs, create tables, set permissions, or configure CORS in the console, load `references/console-ui-guide.md`. 5. **Pre-flight** — Confirm `.catalystrc` and `catalyst.json` exist. If missing, use MCP to get org/project IDs and run `catalyst init --org <orgId> -p <projectId> -ni`. Never use interactive `catalyst init`. 6. **Project structure** — Load `references/project-basics.md` for directory layout, `catalyst.json`, IDs, and dev-to-prod checklist. 7. **CLI questions** — Load `references/cli.md` for the exact command, flags, and safety rules. 8. **Answer** — Provide the specific ID path or CLI command needed. Never ask the user to manually look up IDs when MCP is connected. ## Triggers Use this skill for: "how do I start a Catalyst project", "what is .catalystrc", "where do I find my Table ID / ZAID / Segment ID / Project ID", "difference between Development and Production", "Catalyst project structure", "catalyst.json explained", `catalyst init`, `catalyst deploy`, `catalyst serve`, `catalyst login`, "how do I set up Catalyst with Claude", "how do I use Catalyst in Cursor", "how do I use Catalyst with GitHub Copilot", "which service should I use", "what Catalyst service for X", or any question about Catalyst IDs, environments, organizations, IDE setup, architecture decisions, or CLI subcommands. ## MCP Connection Check (Do This Before Anything Else) **Before reading `.catalystrc`, asking the user for IDs, or doing anything project-related — check whether Zoho MCP tools are available in your tool list.** Look for tools prefixed with `CatalystbyZoho_` (e.g., `CatalystbyZoho_List_All_Projects`, `CatalystbyZoho_List_All_Tables`). **If MCP tools ARE available:** - Use `CatalystbyZoho_List_All_Organizations` to get the org ID - Use `CatalystbyZoho_List_All_Projects` to get the project ID - Use MCP tools to fetch table names, bucket names, ZAIDs, and all other project details directly — do NOT ask the user to copy-paste IDs from the console **If MCP tools are NOT available:** - Prompt the user to connect Zoho MCP before proceeding - Guide them to: **VS Code → Settings → MCP** (or Claude Desktop `claude_desktop_config.json`) and add the Catalyst MCP server - Fall back to reading `.catalystrc` and `catalyst.json` from the local project directory only as a last resort > **Never ask the user to manually look up IDs from the console** if MCP is connected. Every project detail — org ID, project ID, table IDs, ZAIDs, bucket names — is retrievable via MCP tools. Asking the user to hunt for IDs when MCP is available wastes time and introduces copy-paste errors. ## References Load the relevant reference file for detailed information: | Reference | Contents | |-----------|----------| | `references/quick-start.md` | First-time setup — install CLI, `catalyst init`, find org/project IDs, add a function, serve locally, deploy, create a Data Store table, configure CORS/Authorized Domains | | `references/console-ui-guide.md` | Console navigation — finding Project ID/ZAID/Table ID, creating tables with typed columns, enabling App User permissions (Scopes and Permissions), Authorized Domains + CORS toggle | | `references/project-basics.md` | Project directory structure, `catalyst.json`, `.catalystrc`, `catalyst-config.json`, environments, dev-to-prod checklist, all Catalyst IDs (Project ID, ZAID, Table ID, Segment ID, etc.) | | `references/cli.md` | Full CLI command reference — all subcommands with flags, Slate/AppSail non-interactive setup, `catalyst serve` port behavior, deploy scoping, safety rules, resource-first development order | | `references/architecture.md` | Service selection guide — which Catalyst service to use for which pattern, typical stack combinations, DC availability table for regionally restricted services | | `references/setup/claude-code.md` | Installing skills and connecting Zoho MCP in Claude Code (Claude Desktop) | | `references/setup/cursor.md` | Installing skills and connecting Zoho MCP in Cursor | | `references/setup/github-copilot.md` | Installing skills and connecting Zoho MCP in GitHub Copilot (VS Code) |
© catalystbyzoho · Apache-2.0 · catalystbyzoho/agent-skills