executing-planslisted

# Executing Plans ## Overview Load plan, review critically, execute all tasks, report when complete. **Announce at start:** "I'm using the executing-plans skill to implement this plan." **Note:** This path is a legitimate choice, not a fallback. What decides between the two is the plan's size and whether its progress has to outlive the session — inline when the plan finishes in one sitting, superpowersplus:subagent-driven-development when it will not; superpowersplus:writing-plans makes that offer with the measured difference and writes the answer into the plan header's `**Execution:**` field. Where the harness has no subagents (Claude Code, Codex CLI, Codex App, Copilot CLI, and Gemini CLI all have them; see the per-platform tool refs in `../using-superpowers/references/`), there was no choice to present and this is the path. **Progress reports.** Report to your human partner at four fixed points, one line each: - **Starting:** how many tasks the plan has, that this is the inline path, and that the record is this session's todo list. - **Each task done:** `task N of M complete — <short title>`, and what comes next. - **Entering a fix round** — a verification that failed, an audit that came back FAIL, a review finding: which round, out of the cap of three in Step 3, and what failed in a few words. There is one cap on this path and one report; the subagent path has two loops with two caps, and its rounds are not these. - **Finishing:** what was delivered, whi