Skip to content
Back to skills

Catchup

ASecurity

Quick status brief so the user can pick up where things left off — what changed, what state it's in, what's next. Trigger when the user says "catch up", "catch me up", "where were we", "what's the status", "what's going on", or returns to a task/repo after a gap and wants an overview. Output: since last time / current state / next step. Max 12 lines. Grounded in actual git status/log/diff, not guesses.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 19, 2026
ai-agentsgogitapi

Works with

  • api

Security analysis

A100/100

Scanned October 19, 2026

npx -y skills add Jsnnmsc/10000x-engineer --skill catchup --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Catchup?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Catchup
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/jsnnmsc-catchup/badge)](https://www.skillsdirectory.com/skills/jsnnmsc-catchup)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: catchup
description: |
  Quick status brief so the user can pick up where things left off — what changed, what state it's in, what's next.

  Trigger when the user says "catch up", "catch me up", "where were we", "what's the status", "what's going on", or returns to a task/repo after a gap and wants an overview.

  Output: since last time / current state / next step. Max 12 lines. Grounded in actual git status/log/diff, not guesses.
disable-model-invocation: true
---

# Catch Up

Goal: user reads this once and knows exactly what happened, what state things are in, and what to do next — no need to dig through `git log` themselves.

## Rules

- Respond in the same language the user wrote in. If they write in Chinese, reply in Chinese. Technical terms (JWT, Redis, API…) stay in English.
- Max 12 lines. Hard limit.
- Ground every line in evidence: check `git status`, `git log` (recent commits, last ~5-10 or from any supplied branch/date/commit), and `git diff` for uncommitted work before writing anything. Never guess at what changed.
- "Since last time" = what actually landed: commits (with short hash), files changed, uncommitted work in progress. Cite specifics, not "some fixes were made".
- "Current state" = is it clean, mid-change, broken, blocked? Uncommitted changes, failing state, or half-finished work goes here explicitly — don't bury it.
- "Next step" is the closing line and takes the inline form `**Next step:** …` — one obvious action, not a plan. If a blocker is in the way (merge conflict, failing test, unresolved decision), that's the next step instead.
- If there's no git repo or no meaningful history, say so directly rather than padding the response.
- No emoji.

## Output format

```
**Since last time**
- `[hash]` [commit subject]
- Uncommitted: [file] ([new / +N lines / deleted])

**Current state**
[clean / mid-change / blocked — one line, concrete]

**Next step:** [the one obvious action, or the blocker in the way]
```

## Example

`/catchup`

```
**Since last time**
- `6efbe87` Add clear-view and doubt skills, reviewer agent
- `e6fd00f` Bump version to 1.4.0
- Uncommitted: `skills/one-line/SKILL.md` (new, untracked), `README.md` (+12 lines)

**Current state**
Mid-change — new skill written and documented, nothing committed yet; no tests in this repo to run

**Next step:** commit the one-line skill plus the README and version bump as one change
```

## Anti-examples

Don't summarize commits into a theme ("various skill improvements"). The subject lines are
already short — quote them.

Don't report "Current state: clean" when there's untracked work sitting there. Untracked
counts.

Don't turn the next step into a roadmap. One action. If the user needs the plan, they'll ask
for it.

If a branch, date, or "since <commit>" was supplied, scope the git history to that instead of just recent commits.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.