plugins/claude-code-expert/skills/ultraplan/SKILL.md
Ultraplan cloud planning — kick off a plan in the cloud from your terminal, review and revise in the browser, then execute remotely or send back to CLI
npx skillsauth add markus41/claude ultraplanInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Ultraplan offloads plan mode to the cloud. Claude drafts the plan in a Claude Code web session while your terminal stays free. You review sections in your browser, comment, request revisions, then choose to execute remotely or send it back to your CLI.
Available since v2.1.92 (research preview). As of v2.1.101, the first run auto-creates a default cloud environment — no web setup step required.
| Scenario | Use Ultraplan | Use Local Plan Mode | |----------|:---:|:---:| | Complex multi-service migration | ✓ | | | Need to share plan with teammates | ✓ | | | Terminal must stay free during planning | ✓ | | | Quick single-file refactor | | ✓ | | No internet / air-gapped environment | | ✓ | | Iterating rapidly on small tasks | | ✓ |
/ultraplan migrate the auth service from sessions to JWTs
Include the keyword "ultraplan" in any prompt:
> ultraplan: design the database migration strategy for the multi-tenant switch
The browser interface shows the plan as structured sections:
Plan history is preserved — you can compare revisions and revert.
Claude Code on the web runs the implementation. Results and changed files are available in the session. Push from the web session or download a patch.
The approved plan is serialized and sent back to your local terminal session. Claude continues from there with the agreed plan as context.
After executing in the cloud and pushing a branch, use /autofix-pr to hand the PR over to Claude:
> /autofix-pr
Claude detects the open PR for your current branch and enables auto-fix — watches CI, fixes failures, handles review comments, pushes until green.
Ultraplan uses Opus (opus → latest claude-opus-4-8) by default for the planning phase — the reasoning depth justifies the cost for complex migrations and architectural decisions. Pair it with a high effort level (xhigh/max on Opus 4.8) for the hardest plans. The execution phase uses the model appropriate to the task.
Estimated cost for a complex plan: ~$0.50–$2.00 depending on context volume. Simple plans use fewer tokens and run faster.
Use ultraplan to design a feature, then run /team-onboarding to generate a ramp-up guide that explains the resulting architecture to new teammates:
> /team-onboarding
This generates a guide from your local Claude Code usage patterns and the project structure — useful after landing a large feature.
development
Enhanced plan-authoring skill with Pre-Writing context gathering, task metadata, non-TDD templates, Red Flags, telemetry, and an automated plan linter. Use when you have a spec or requirements for a multi-step task, before touching code.
tools
Documentation intelligence engine with graph-based API docs, algorithm library, and drift detection
tools
--- name: mcp description: Configure MCP servers for Claude Code — stdio vs HTTP, authentication, Tools/Resources/Prompts distinction, channels (CI webhook, mobile relay, Discord bridge, fakechat), and cost of always-loaded tools. Use this skill whenever adding an MCP server, debugging connection issues, choosing between MCP Tools vs Prompts vs Resources, installing channel servers, or managing .mcp.json. Triggers on: "MCP server", "mcp config", "add Obsidian MCP", "install context7", "channels"
tools
Design, install, and debug Claude Code hooks across the full lifecycle (PreToolUse, PostToolUse, PostToolUseFailure, UserPromptSubmit, Notification, Stop, SessionStart, SessionEnd, PreCompact, SubagentStart, SubagentStop, TeammateIdle, PermissionRequest, Setup). Use this skill whenever a user asks to "install hooks", "add a pre-tool hook", "format on save", "block dangerous commands", "protect sensitive files", "restore context after compact", "enforce tests before stop", capture subagent telemetry, or runs /cc-hooks. Also triggers on "hooks not firing", "hook keeps blocking", or any configuration of .claude/settings.json hook sections.