Calavera

Calavera

Use Calavera to compose and apply project tooling through its MCP tools whenever they are available. This skill guides the active agent; it is not a subagent by itself.

Start Here

1. Confirm whether Calavera MCP tools are available. Look for tools named inspect_project, list_profiles, list_integrations, list_ai_artifacts, compose_recipe, validate_recipe, explain_recipe, dry_run_apply, and apply_recipe.
2. If the Calavera MCP tools are not available, stop and help the user register or repair the MCP server before composing or applying anything. Do not inspect npm cache internals or import Calavera source files from package cache paths as a substitute for MCP setup.
3. Call inspect_project when available, or inspect the project manually only when the MCP server cannot be registered and the user chooses a fallback flow. Check files such as package.json, calavera.config.json, .editorconfig, eslint.config.js, oxlint.json, .prettierrc.json, .stylelintrc.json, and tsconfig.json. If likely conflicts exist, pause before applying changes. List each conflict as a hard stop or a migration decision the user can approve, and use dry_run_apply to show concrete impact when adoption still looks possible.
4. Use AskUserTool or the agent client's equivalent when available to clarify profile preferences, framework needs, conflict decisions, and apply approval. If no such tool exists, ask the user directly.
5. List choices with list_profiles, list_integrations, and list_ai_artifacts. Use describe_integration when the user asks for more information or when you need to compare options.
6. Choose either Oxfmt or Prettier for formatting, never both.
7. Once the profile and requirements are clear, compose the recipe with compose_recipe.
8. Validate and explain it with validate_recipe and explain_recipe.
9. Present dry_run_apply output to the user before changing files, including inspection findings, omitted script explanations, ownership notes, and planned file changes.
10. Call apply_recipe only after the user explicitly approves the dry run.
11. If the MCP transport closes or reports -32000 during or immediately after apply_recipe, treat the outcome as unknown instead of failed. Inspect calavera.config.json, .calavera/state.json, generated files, and package metadata before retrying the apply.

Do not hand-author calavera.config.json when the Calavera MCP server is available. Let Calavera compose, validate, dry-run, and apply the recipe so generated files, package scripts, dependencies, AI artifacts, and managed state stay consistent.

Treat files listed by dry_run_apply as Calavera-managed outputs. Do not hand-write or edit them; let apply_recipe or create-project-calavera apply create them after approval.

apply_recipe.writeConfig: false only skips writing calavera.config.json. Do not use it to bypass managed-file conflicts, stale state hashes, or an unapproved dry-run result.

Use reownManagedFiles only after the user explicitly approves treating listed Calavera-tracked files' current contents as intentional local state before dry-run/apply.

MCP Setup

If the Calavera MCP tools are not available, help the user register this server from the project root. Check the target project's package.json first and use the package manager declared by packageManager or devEngines.packageManager.

{
  "mcpServers": {
    "calavera": {
      "command": "npx",
      "args": ["--package", "create-project-calavera@<version>", "create-project-calavera-mcp"]
    }
  }
}

For manual MCP setup, use npx --package create-project-calavera@<version> create-project-calavera-mcp for npm-managed projects, pnpm dlx --package create-project-calavera@<version> create-project-calavera-mcp for pnpm, yarn dlx --package create-project-calavera@<version> create-project-calavera-mcp for Yarn, and bunx --package create-project-calavera@<version> create-project-calavera-mcp for Bun. In JSON-based MCP configs, put the first word in command and the remaining words in args. Matching the project package manager prevents package-manager preflight failures before Calavera can start, such as npm rejecting a Bun-managed project. Keep an explicit version so package-manager launchers resolve the create-project-calavera-mcp bin reliably without making the persistent MCP registration float to a later package release.

After registering the MCP server, reload or restart the agent session if the MCP host does not discover new tools dynamically. Confirm the Calavera tools are visible before composing a recipe.

npm create needs -- before Calavera flags, for example npm create project-calavera -- --init. Direct binary launchers such as npx --package create-project-calavera@<version> create-project-calavera --help do not need an extra -- before Calavera flags. Avoid npx --package create-project-calavera@<version> create-project-calavera -- --help; MCP registrations launch create-project-calavera-mcp directly and should not add --help.

If a Bun-based MCP launch fails before Calavera starts with error: bun is unable to write files to tempdir: PermissionDenied, configure the MCP host to set TMPDIR to an absolute writable directory for that server. If Bun's package cache is also restricted, set BUN_INSTALL_CACHE_DIR to an absolute writable cache directory. Keep these overrides on Bun registrations only; they are recovery settings for restricted hosts, not default Calavera MCP config.

If this project was bootstrapped with create-project-calavera --init, also check .agents/calavera/mcp.md for local setup notes.

For Claude Code, prefer a project-scoped .mcp.json in the project root when the team should share the Calavera server registration. Do not put the server under .claude/settings.json; Claude Code does not load MCP servers from that file. You can also use claude mcp add to register the same package-manager-specific command.

Registering the Calavera MCP server is a persistent code-execution change that runs an external package, so ask for explicit user approval before creating .mcp.json, running claude mcp add, or approving the first server launch.

Fallbacks

If the MCP server cannot be registered, use the hosted Calavera Web UI to compose and download a recipe:

https://calavera.schalkneethling.com

After a recipe exists, preview local changes with the package-manager-specific apply dry-run command, pinned to the same explicit package version used for MCP setup: npm create project-calavera@<version> apply -- --dry-run for npm, pnpm dlx create-project-calavera@<version> apply --dry-run for pnpm, yarn dlx create-project-calavera@<version> apply --dry-run for Yarn, or bunx create-project-calavera@<version> apply --dry-run for Bun. Ask the user to approve the preview before running the matching apply command.

User Prompt

When the user wants to start from a scaffolded or existing project, suggest:

Use Calavera for this project. First verify that the Calavera MCP tools are available. If they are not available, stop and help me configure the MCP server before composing or applying anything. Once the tools are available, inspect the current project for existing tooling and possible config conflicts, list the available profiles, integrations, and AI artifacts, compose a recipe, show me the dry-run result, and apply it only after I approve.