Update to AgentGrid.io - Shared Drive

{
  "name": "agent-grid-sandboxed",
  "description": "Create, inspect, review, publish, and edit Agent Grid artifacts entirely through MCP in a sandboxed agent.",
  "included_files": [],
  "skill_md_contents": "---\nname: agent-grid-sandboxed\ndescription: \"Create, inspect, review, publish, and edit Agent Grid artifacts entirely through MCP in a sandboxed agent.\"\nmcpServers:\n  - anima\ncompatibility: \"Use this skill without a shell, Git, GitHub, or direct internet uploads.\"\nhomepage: \"https://github.com/AnimaApp/mcp-server-guide\"\nmetadata: {\"clawdbot\":{\"emoji\":\"🎨\",\"requires\":{\"env\":[\"ANIMA_API_TOKEN\"]},\"primaryEnv\":\"ANIMA_API_TOKEN\"},\"author\":\"animaapp\",\"version\":\"2.0.0\"}\n---\n\n<!-- Generated from skills/agent-grid-sandboxed/SKILL.md by scripts/build-mirror.sh — do not edit directly. -->\n\n# Use Agent Grid in a sandbox\n\nAgent Grid is a governed space where humans and agents share live artifacts. Each artifact is one repository, filed in one of your team's workspaces.\n\nYour access is limited to what the human who connected you consented to. It can be revoked at any time.\n\nUse this skill when you cannot use a shell, Git, GitHub, or direct internet uploads.\n\n## Create an artifact\n\nCreate all code first. Then call `artifact-create` with `files`. This sandboxed tool accepts only inline file creation; it does not take `type`.\n\nUse `files` only for UTF-8 text files. The map key is the file path.\n\nDo not call `artifact-create` to change an existing artifact. Each call creates a new artifact.\n\nExample:\n\n```\nartifact-create(\n  files: {\n    \"index.html\": \"<h1>Status</h1>\",\n    \"styles.css\": \"h1 { color: navy; }\"\n  },\n  name: \"Status page\"\n)\n```\n\nThe call returns `sessionId`, `revision`, `artifactUrl`, and the applicable preview URLs. The artifact is ready immediately. Share `artifactUrl` with the human.\n\nTo create a knowledge artifact from Anima's template, call `artifact-create_knowledge()` or pass only an optional `name`. Do not pass files, a framework, or an artifact type.\n\n## Change artifact code\n\nUse `artifact-explore` to list, search, read, or review artifact content.\n\nUse the returned `revision` as `baseRevision` in `artifact-edit`. The edit creates one commit.\n\nUse `str_replace`, `write`, `delete`, or `move`. Do not use any external upload process.\n\nRead [MCP edit workflow](https://github.com/AnimaApp/mcp-server-guide/blob/main/skills/agent-grid-sandboxed/references/edit-workflow.md) before you change content.\n\n## Handle reviews\n\nCall `review-list` when a human mentions comments and when you resume reviewed work. Read replies before acting.\n\nPut each addressed comment's `resolveTrailer` in the `artifact-edit` commit message. Use `review-reply` when the conversation must stay open. Use `review-resolve` only when no commit will close the comment.\n\n## Use lifecycle tools\n\n- Use `workspace-list_artifacts()` to find recent readable artifacts. The result is bounded, so a short or empty list is not proof that the workspace has no other artifacts.\n- Use `artifact-update_metadata` to change the name or privacy.\n- Use `artifact-duplicate` to create an independent copy.\n- Use `artifact-publish` only when the user asks for a public deployment.\n- Use `artifact-unpublish` to stop the public deployment.\n- Use `artifact-delete` only when the user explicitly asks for deletion.\n\nRead [MCP tools](https://github.com/AnimaApp/mcp-server-guide/blob/main/skills/agent-grid-sandboxed/references/mcp-tools.md) for exact parameters. Read [workflows](https://github.com/AnimaApp/mcp-server-guide/blob/main/skills/agent-grid-sandboxed/references/workflows.md) for sequences.\n\nRead [troubleshooting](https://github.com/AnimaApp/mcp-server-guide/blob/main/skills/agent-grid-sandboxed/references/troubleshooting.md) when an MCP operation fails.\n"
}