← ClaudeAtlas

planlisted

Plan a change into verifiable tasks with acceptance criteria, dependencies and a bounded verification command each — dry-run and explicit confirmation before anything is written. Never implements. Use when the user asks to plan or break down work, or invokes /rein:plan.
luisfarfan/rein-agentic-kit · ★ 0 · AI & Automation · score 73
Install: claude install-skill luisfarfan/rein-agentic-kit
# /rein:plan Thin orchestration for the **planner** role. Turns a request into a task list `/rein:run` can execute without re-deriving intent. **This skill never implements, never ticks a checkbox and never invokes `/rein:run`.** Usage: `/rein:plan <change-name> [description]` ```bash R=$(command -v rein || ls -d ~/.claude/plugins/cache/*/rein/*/bin/rein 2>/dev/null | sort -V | tail -1) ``` ## Steps 1. **Record this invocation** — never blocks, never fails the run. Shell state does NOT persist between tool calls, so `R` is resolved and used in the SAME block or it is empty and nothing is recorded: ```bash R=$(command -v rein || ls -d ~/.claude/plugins/cache/*/rein/*/bin/rein 2>/dev/null | sort -V | tail -1); "$R" event plan ``` 2. Resolve the project so the plan lands where the loop will look for it and the verification commands are real: `"$R" context .` — note `plan.source`, `plan.path`, `config.commands` and `verifyPolicy.mode`. 3. Understand the request before decomposing it. Use bounded search and precise reads, not whole files. If scope, an interface or a trade-off is genuinely ambiguous, **ask now** — a wrong assumption here is paid for by every implementation step after it. 4. Draft the plan. It opens with a **short header** and then the tasks: ```markdown # Change: add-widget ## Why One or two sentences: what problem this solves, and what happens if it is not done. This is what the reviewer judges the work against.