.cursor/skills/plan-executor/SKILL.md
Orchestrates execution of Cursor Plans with multiple TODOs by spawning one todo-completer subagent per TODO in serial order, managing plan state, git commits, and human escalation. TRIGGER WHEN: User asks to execute a plan that has more than 3 TODOs in it
npx skillsauth add Huddo121/My-Agent-Loop plan-executorInstall 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.
Orchestrate execution of plans from .cursor/plans/*.plan.md by delegating each TODO to a todo-completer subagent, maintaining plan state, and handling git commits between completed todos.
Apply this skill when the user asks to execute a plan and that plan has more than 3 TODOs. For plans with a few TODOs, execute directly without orchestration.
Before starting plan execution:
git branch --show-current. If on main or master, you must not push there. Checkout a new branch (e.g. git checkout -b plan-<plan-name-slug>)..cursor/plans/*.plan.md. The plan frontmatter contains todos as an ordered list of { id, content, status } where status is pending, in-progress, or completed.Execute TODOs strictly in order. One subagent per TODO. No batching or parallelisation.
For each pending TODO:
Set the TODO's status to in-progress in the plan file's frontmatter.
Invoke the todo-completer subagent with:
Do not assume how the subagent is triggered (mcp_task, agent reference, etc.) – use whatever mechanism your harness provides.
Success
status to completed in the plan fileFailure (task completion)
Failure (infrastructure)
| Rule | Why | |------|-----| | No parallel subagents | Serial execution only | | Never push to main/master | Must work on a feature branch | | Parent owns plan file state | Subagent does not update todo status | | Parent owns git commits | Add, commit, push after each successful TODO | | No code quality checks between TODOs | Codebase may be intentionally broken mid-plan, sub-agents are responsible for this check | | Subagent owns code quality at end of its run | Verification is the subagent's job |
If a subagent reports that it expected the codebase to be "working" (e.g. typecheck passes, tests pass) but it is not, and that blocks progress: ask for human input. Do not spawn another subagent hoping it will fix the intermediate state.
Plans use YAML frontmatter:
---
name: Plan Name
overview: Brief description
todos:
- id: todo-1
content: Description of work
status: pending # pending | in-progress | completed
- id: todo-2
content: ...
status: pending
isProject: false
---
When updating status, preserve the rest of the frontmatter and only change the relevant TODO's status field.
development
Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI. Provides project context, component docs, and usage examples. Applies when working with shadcn/ui, component registries, presets, --preset codes, or any project with a components.json file. Also triggers for "shadcn init", "create an app with --preset", or "switch to --preset".
development
Guides agents through autonomous codebase exploration, deep questioning, and upfront decision-making to produce self-contained Cursor Plans for feature work. TRIGGER WHEN: The user asks to plan, design, or scope a feature or change before implementation, especially when a less capable agent will execute the plan.
development
Reviews code changes in this repository for correctness, edge cases, and alignment with local patterns. Use when the user asks for a code review of diffs, pull requests, or modified files in this project.
data-ai
Example TaskFlow authoring pattern for inbox triage. Use when messages need different treatment based on intent, with some routes notifying immediately, some waiting on outside answers, and others rolling into a later summary.