Course-correct mid-session and reassess priorities. Triggered by 'zoom out', 'step back', 'reassess', 'am I on track', 'am I solving the right problem', or checking work against real goals.
Scanned 9/2/2026
Install to Claude Code
npx -y skills add tomcounsell/ai --skill zoom-out --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Zoom Out?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/tomcounsell-zoom-out)More formats (shields.io, HTML) on the badges page.
---
name: zoom-out
description: "Course-correct mid-session and reassess priorities. Triggered by 'zoom out', 'step back', 'reassess', 'am I on track', 'am I solving the right problem', or checking work against real goals."
allowed-tools: Read, Bash
---
# Skill: /zoom-out
## Repo Context Probe
If `.claude/skill-context/zoom-out.md` exists, read it and honor its declarations; otherwise use the generic defaults described below.
The context file is where a repo declares project-local tools for the steps below — a memory/recall CLI to synthesize recent context (Step 1) and a messaging CLI for remote delivery of the summary (Step 5). When the file is absent (the common case in a foreign repo), reconstruct context from `git log` and any session notes, and just print the summary in-session.
## Purpose
Pull back from the current task, reassess priorities, and reorient toward the actual goal — before more time is spent solving the wrong problem.
## When to Use
- You are on a third consecutive patch loop on the same issue and it is still not fixed
- The current sprint has drifted from the original goal
- A session has gone quiet or circular — same fixes, same failures
- The user says "step back", "zoom out", or "are we doing the right thing?"
- Before starting a fourth implementation attempt on something that keeps breaking
## Steps
1. **Synthesize recent context.** If the repo context file declares a memory/recall CLI, use it to surface observations related to the current work area and read the top results, noting any corrections or pattern observations that apply. Otherwise reconstruct recent context from `git log` (Step 3) and any session notes — there is no generic memory store to query.
2. **Check open GitHub issues.** Get a snapshot of active work:
```bash
gh issue list --state open --limit 10
```
Note which issues are in progress, which are blocked, and which have been open the longest.
3. **Read recent session context.** If a plan doc or PROGRESS.md is available at the worktree root, read it. Check the last 10 git log entries:
```bash
git log --oneline -10
```
4. **Produce a strategic summary.** Write a concise assessment covering:
- **What we set out to do** (original goal)
- **What we've actually done** (last 5 actions/commits)
- **Where we are stuck** (current blocker, if any)
- **Recommended next focus** (highest-leverage action)
- **Deprioritization list** (2–3 things that can wait)
5. **Deliver the summary.** Print it in the session. If the user is remote and the repo context file declares a messaging CLI, optionally send the summary through it.
6. **Ask one question.** Close with: "Does this match your mental model, or is there something I'm missing?"
## Output
A strategic summary with recommended next focus and deprioritization list. One follow-up question.
## Anti-Patterns
- Do not use /zoom-out as a stalling tactic — if you know what to do next, do it.
- Do not run /zoom-out on every session — it is for course correction, not routine check-ins.
- Do not produce a long report — the summary should fit in a Telegram message (< 300 words).
- Do not skip the "recommended next focus" — a zoom-out without a recommendation is just a status report.
- Do not re-read all files from scratch — use memory search and git log to reconstruct context efficiently.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!