# Import Existing Ai Config > Import existing AI tool configurations (from Claude, Copilot, Cursor, or Codex) into universal-ai-config templates. Converts target-specific files into universal templates. - **Type:** Skill - **Install:** `agentstack add skill-fabis94-universal-ai-config-import-existing-ai-config` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [fabis94](https://agentstack.voostack.com/s/fabis94) - **Installs:** 0 - **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [fabis94](https://github.com/fabis94) - **Source:** https://github.com/fabis94/universal-ai-config/tree/main/.universal-ai-config/skills/import-existing-ai-config ## Install ```sh agentstack add skill-fabis94-universal-ai-config-import-existing-ai-config ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # Import Existing AI Config Convert existing target-specific AI configuration files into universal-ai-config templates. The user must specify which target to import from. ## Usage The user should specify the source target: `claude`, `copilot`, `cursor`, or `codex`. ## Import Process ### 1. Identify Source Files Scan the target's config directory for existing configuration files: **Claude** (`.claude/`): - Instructions: `.claude/rules/*.md` — each file has `description` and optional `paths` frontmatter - Skills: `.claude/skills/*/SKILL.md` — skill directories with frontmatter (`name`, `description`, `allowed-tools`, `model`, `context`, `agent`, `disable-model-invocation`, `user-invocable`, `argument-hint`, `hooks`). Skill directories may also contain extra supporting files (references, examples, scripts) alongside `SKILL.md`. - Agents: `.claude/agents/*.md` — agent files with frontmatter (`name`, `description`, `tools`, `disallowedTools`, `permissionMode`, `skills`, `hooks`, `memory`, `model`) - Hooks: `.claude/settings.json` → `hooks` key — JSON with PascalCase event names (`SessionStart`, `SessionEnd`, `UserPromptSubmit`, `PreToolUse`, `PostToolUse`, `PostToolUseFailure`, `Stop`, `SubagentStart`, `SubagentStop`, `PreCompact`, `PermissionRequest`, `Notification`) - MCP: `.mcp.json` — JSON with `mcpServers` wrapper containing server configs (`type`, `command`, `args`, `env`, `url`, `headers`) - Commands (deprecated): `.claude/commands/*.md` — single-file slash commands with optional frontmatter (`description`, `allowed-tools`, `argument-hint`, `model`). May include subdirectories for namespacing (e.g. `.claude/commands/frontend/component.md` — command name is `component`). Body may use `$ARGUMENTS` placeholder, `!` prefix for bash execution, and `@` prefix for file references. **Copilot** (`.github/`): - Instructions: `.github/copilot-instructions.md` (always-apply) and `.github/instructions/*.instructions.md` (with `applyTo` frontmatter) - Skills: `.github/skills/*/SKILL.md` — skill directories with frontmatter (`name`, `description`, `license`, `compatibility`, `metadata`). May contain extra supporting files alongside `SKILL.md`. - Agents: `.github/agents/*.agent.md` — agent files with frontmatter (`name`, `description`, `tools`, `model`, `target`, `mcp-servers`, `handoffs`) - Hooks: `.github/hooks/hooks.json` — JSON with version field and camelCase event names (`sessionStart`, `sessionEnd`, `userPromptSubmitted`, `preToolUse`, `postToolUse`, `errorOccurred`) - MCP: `.vscode/mcp.json` — JSON with `servers` wrapper (not `mcpServers`), may include `inputs` array for interactive secret prompts **Cursor** (`.cursor/`): - Instructions: `.cursor/rules/*.mdc` or `.cursor/rules/*.md` — with `description`, `globs`, `alwaysApply` frontmatter - Skills: `.cursor/skills/*/SKILL.md` — skill directories with frontmatter (`name`, `description`, `license`, `compatibility`, `metadata`, `disable-model-invocation`). May contain extra supporting files alongside `SKILL.md`. - Hooks: `.cursor/hooks.json` — JSON with version field and camelCase event names (`sessionStart`, `sessionEnd`, `beforeSubmitPrompt`, `preToolUse`, `postToolUse`, `postToolUseFailure`, `stop`, `subagentStart`, `subagentStop`, `preCompact`, plus Cursor-specific events like `beforeShellExecution`, `afterFileEdit`) - MCP: `.cursor/mcp.json` — JSON with `mcpServers` wrapper, omits `type` field (Cursor infers transport from `command` vs `url`) - Note: Cursor does not have agents **Codex** (mix of root + `.codex/` + `.agents/`): - Instructions: `AGENTS.md` at project root (free-form markdown, no frontmatter — concatenation of always-on guidance) and `/AGENTS.override.md` at nested directories (per-directory override files). Split into separate universal instructions: H2-section split for the root file (one universal instruction per `## heading` block), one universal instruction per AGENTS.override.md file with `globs: ["/**"]`. - Skills: `.agents/skills/*/SKILL.md` (root-relative, the open Agent Skills standard location) — minimal frontmatter (`name`, `description`, `version`, `author`, `license`, `compatibility`, `metadata`). Sidecar `agents/openai.yaml` next to SKILL.md contains UI metadata (`interface.display_name`, etc.), invocation policy (`policy.allow_implicit_invocation`), and MCP dependencies (`dependencies.tools[]`). - Agents: `.codex/agents/*.toml` — standalone TOML files with fields `name`, `description`, `developer_instructions` (body), `model`, `model_reasoning_effort`, `sandbox_mode`, `nickname_candidates`, `mcp_servers`, `skills.config`. - Hooks: `.codex/hooks.json` — JSON with PascalCase event names (`SessionStart`, `UserPromptSubmit`, `PreToolUse`, `PostToolUse`, `PermissionRequest`, `Stop`). Handler fields: `type` (only `command`), `command` (single shell string — args inline), `matcher`, `timeout`, `statusMessage`. - MCP: `.codex/config.toml` `[mcp_servers.]` tables — TOML format with fields `command`, `args`, `env`, `url`, `cwd`, `env_vars`, `http_headers`, `bearer_token_env_var`, `env_http_headers`, `enabled`, `required`, `enabled_tools`, `disabled_tools`, `startup_timeout_sec`, `startup_timeout_ms`, `tool_timeout_sec`, `oauth_resource`, `scopes`, `experimental_environment`. Other top-level keys in `config.toml` (`[profiles.*]`, `[model_providers.*]`, `[permissions.*]`, `personality`, etc.) are user content — preserve them or note them separately; don't import as templates. ### 2. Convert to Universal Format For each file found, convert it to a universal-ai-config template: **Frontmatter mapping** (target-specific → universal): | Claude | Copilot | Cursor | Codex | Universal | | -------------------------- | ------------------------- | -------------------------- | ------------------------------------------------------------------------------------ | ---------------------------- | | `paths` | `applyTo` | `globs` | (derived from AGENTS.override.md dir) | `globs` | | (no paths field) | (copilot-instructions.md) | `alwaysApply: true` | (root AGENTS.md content) | `alwaysApply: true` | | `disable-model-invocation` | — | `disable-model-invocation` | `policy.allow_implicit_invocation: false` (inverted) | `disableAutoInvocation` | | `user-invocable` | — | — | — | `userInvocable` | | `allowed-tools` | — | — | — | `allowedTools` | | `context: fork` | — | — | — | `forkContext: true` | | `agent` | — | — | — | `subagentType` | | `argument-hint` | — | — | — | `argumentHint` | | `hooks` | — | — | — | `hooks` | | (commands: manual-only) | — | — | — | `disableAutoInvocation` | | — | `excludeAgent` | — | — | `excludeAgent` | | — | `license` | `license` | `license` | `license` | | — | `compatibility` | `compatibility` | `compatibility` | `compatibility` | | — | `metadata` | `metadata` | `metadata` | `metadata` | | — | `target` | — | — | `target` | | — | `mcp-servers` | — | `mcp_servers` (agent TOML) | `mcpServers` | | — | `handoffs` | — | — | `handoffs` | | — | — | — | `version` (SKILL.md) | `version` | | — | — | — | `author` (SKILL.md) | `author` | | — | — | — | `interface.*` (openai.yaml) | `codex.interface.*` | | — | — | — | `dependencies.tools[]` (openai.yaml) | `codex.dependencies.tools[]` | | — | — | — | `nickname_candidates` (agent TOML) | `nicknameCandidates` | | — | — | — | `sandbox_mode` (agent TOML) | `sandboxMode` | | `effort` | — | — | `model_reasoning_effort` (agent TOML, when value in {minimal,low,medium,high,xhigh}) | `effort` | **Hook event mapping** (target-specific → universal): | Claude | Copilot | Cursor | Codex | Universal | | -------------------- | --------------------- | -------------------- | ------------------- | -------------------- | | `SessionStart` | `sessionStart` | `sessionStart` | `SessionStart` | `sessionStart` | | `SessionEnd` | `sessionEnd` | `sessionEnd` | — | `sessionEnd` | | `UserPromptSubmit` | `userPromptSubmitted` | `beforeSubmitPrompt` | `UserPromptSubmit` | `userPromptSubmit` | | `PreToolUse` | `preToolUse` | `preToolUse` | `PreToolUse` | `preToolUse` | | `PostToolUse` | `postToolUse` | `postToolUse` | `PostToolUse` | `postToolUse` | | `PostToolUseFailure` | — | `postToolUseFailure` | — | `postToolUseFailure` | | `Stop` | — | `stop` | `Stop` | `stop` | | `SubagentStart` | — | `subagentStart` | — | `subagentStart` | | `SubagentStop` | — | `subagentStop` | — | `subagentStop` | | `PreCompact` | — | `preCompact` | — | `preCompact` | | `PermissionRequest` | — | — | `PermissionRequest` | `permissionRequest` | | `Notification` | — | — | — | `notification` | | — | `errorOccurred` | — | — | `errorOccurred` | Cursor-specific events (e.g. `beforeShellExecution`, `afterFileEdit`) should be preserved as-is — they pass through to Cursor and are dropped for other targets. **Hook handler field mapping:** | Claude | Copilot | Cursor | Universal | | --------------------------- | ------------ | --------- | --------- | | `command` | `bash` | `command` | `command` | | `timeout` | `timeoutSec` | `timeout` | `timeout` | | `matcher` (in parent group) | — | `matcher` | `matcher` | **MCP server conversion** (target-specific → universal): The universal MCP format uses `mcpServers` as the wrapper key, with each server having `type`, `command`, `args`, `env`, `url`, and `headers` fields plus any of the Codex-specific extensions documented in ``. | Source | Conversion | | -------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Claude `.mcp.json` | Copy `mcpServers` as-is (already matches universal format) | | Copilot `.vscode/mcp.json` | Rename `servers` → `mcpServers`, add `type` field if missing, copy `inputs` … ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [fabis94](https://github.com/fabis94) - **Source:** [fabis94/universal-ai-config](https://github.com/fabis94/universal-ai-config) - **License:** MIT Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-fabis94-universal-ai-config-import-existing-ai-config - Seller: https://agentstack.voostack.com/s/fabis94 - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.