Generate TDD implementation plans for zero-context engineers

Writing Plans produces implementation plans written for an engineer who is skilled but has zero context on the codebase, toolset, or problem domain and may not know good test design - so every plan documents exactly which files to touch, complete code rather than "add validation," exact test commands with expected output, and how to verify each step, broken into bite-sized tasks. It announces itself at the start, runs inside a dedicated worktree created by the brainstorming skill, and saves the plan to docs/plans/YYYY-MM-DD-<feature-name>.md. Every plan opens with a required header naming the required superpowers:executing-plans sub-skill, a one-sentence Goal, a 2-3 sentence Architecture summary, and the Tech Stack. Each task lists exact file paths to create, modify with line ranges, or test, then breaks the work into 2-5 minute TDD steps: write the failing test, run it to confirm it fails, write the minimal implementation, run it to confirm it passes, and commit - each step given as a literal command with its expected output, following DRY, YAGNI, TDD, and frequent-commit principles. After saving, it offers the user an explicit choice between two execution paths: subagent-driven development in the current session, a fresh subagent dispatched per task with review between tasks via the required superpowers:subagent-driven-development sub-skill, or a parallel session opened separately that uses superpowers:executing-plans for batch execution with checkpoints.

Use it to turn a design or feature idea into a plan detailed enough for an engineer, or an AI agent, with no prior context to execute correctly, when TDD-style, verifiable, bite-sized steps matter more than a quick high-level outline. It's meant to run inside a dedicated worktree already set up by the brainstorming skill, not as a standalone planning exercise outside that flow.

Input: a feature or change to plan, plus the codebase context needed to name exact files and write complete code for each step. Output: a saved markdown plan, docs/plans/YYYY-MM-DD-<feature-name>.md, with a required header, and one task block per component containing exact file paths, complete code, exact commands with expected output, and a commit step, plus, after saving, an explicit execution-path choice for the user.

• brainstorming skill, which creates the worktree this skill runs in
• superpowers:executing-plans, required for a plan's own execution, referenced in the header and for the parallel-session path
• superpowers:subagent-driven-development, required for the subagent-driven execution path

Anyone directing an AI agent, or a genuinely context-free engineer, to implement a feature, who needs a plan detailed enough to execute without guessing at file paths, test design, or commands.