agentleFS
TheVellerTheVeller/claude-skills/skills/engineering/repo-sync/SKILL.md

repo-sync

Skill4 starsChanged 12 days agoMITView on GitHub

Use when the user wants to confirm a local git repo matches its GitHub remote and pull the latest before working — phrasings like "check if local has the latest from GitHub", "is this repo up to date with the remote", "sync this repo", "pull Lovable's changes first", or at the start of local work on a repo that's also edited in the cloud (Lovable / Supabase projects like videomatic). Detects up-to-date / behind / ahead / diverged, fast-forwards when lossless, and can overwrite the local copy with the remote — always backing up first — when they've diverged.

  • Reads credentials
  • Deletes or force-pushes

8 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

repo-sync

Make the local working copy reflect GitHub's latest before you start editing.

Why this exists

Some repos are edited in two places: the cloud (e.g. Lovable committing to GitHub for the videomatic / pixel-orchestrator-ui project) and locally with Claude Code. GitHub holds the newest cloud changes, so the local clone silently drifts behind. Starting local work on a stale clone means editing old code and creating avoidable conflicts. This skill checks the gap first and brings local up to GitHub's state — overwriting local when needed, but never without a recoverable backup.

Tool

A single script does all the git plumbing so you don't re-derive it each time. Run it from inside the target repo (the current working directory's repo). Resolve its path relative to this SKILL.md: scripts/repo-sync.sh.

bash <skill-dir>/scripts/repo-sync.sh check        # READ-ONLY: fetch + report state
bash <skill-dir>/scripts/repo-sync.sh sync         # apply the safe action
bash <skill-dir>/scripts/repo-sync.sh sync --force # also allow overwriting local (with backup)

The script reports state: as one of:

StateMeaningWhat sync does
up-to-datelocal == remotenothing
behindremote has commits local lacksfast-forward (lossless)
aheadlocal has commits not on remoterefuses unless --force; then backs up + hard-resets to remote
divergedboth have unique commitsrefuses unless --force; then backs up + hard-resets to remote

Workflow

  1. Always start with check. It's read-only — it fetches and prints the state, the ahead/behind counts, and a recommended action. Show the user that line.
  2. Decide based on state:
    • up-to-date → nothing to do. Say so and proceed with the real work.
    • behind → run sync. This is a plain fast-forward, lossless. Safe to do without asking (it's exactly what "pull the latest" means). If the tree is dirty the script will ask for --force so it can stash first — mention the stash to the user.
    • ahead or diverged → stop and confirm with the user before overwriting. Local commits exist that aren't on GitHub. The user's intent for this skill is "let GitHub win and start from there," but that discards local commits, so surface exactly what would be lost (git log @{u}..@ --oneline) and confirm. Only then run sync --force.
  3. Report what happened — the resulting state and, for any overwrite, the backup branch name and stash the script created.

Safety — nothing is lost

sync --force never destroys silently. Before any git reset --hard or fast-forward over a dirty tree, the script:

  • creates a backup/pre-sync-<timestamp> branch at the current commit (recovers committed work), and
  • runs git stash push -u (recovers uncommitted and untracked files).

So even a "wrong" overwrite is reversible: git checkout backup/pre-sync-<ts> or git stash list → git stash pop. Tell the user these exist rather than assuming they'll read the script output.

This script wraps git reset --hard deliberately, so the vault's PreToolUse guardrail (which blocks bare git reset --hard) won't fire — the backups are what make that acceptable. Don't work around the guardrail by other means; route destructive syncs through this script so the backup always happens.

Notes for the videomatic repo specifically

  • Remote: pixel-orchestrator-ui.git, branch main, upstream origin/main.
  • It's a standalone git repo nested under the vault; the vault's auto-commit daemon does not commit it, so its uncommitted changes are real local work — don't assume a clean tree.
  • Untracked files ending in 2 (e.g. Editor 2.tsx, .env 2.example) are Google Drive sync-conflict duplicates, not real source. A force-sync stashes them away (recoverable), which is usually the desired cleanup.

After many syncs

Backup branches accumulate. To prune old ones once you're sure they're not needed: git branch --list 'backup/pre-sync-*' then delete the stale ones with git branch -D <name>.

Lane (Grok Bot fleet)

  • Canonical copy lives on Ignacio's Mac (Hermes / Claude Code): ~/.hermes/skills/repo-sync/ and vault .claude/skills/repo-sync/.
  • Overnight or Cloud Agent work that touches a Mac clone → SWE briefs Hermes to run repo-sync first (SWE does not clone on the Grok box).
  • Pure Cloud Agent on the GitHub remote → no local clone; skip this skill (the remote is the source of truth for the agent).
  • Matching Lovable projects ↔ local folders: list Lovable via MCP, list local 02_Programs/3-Software/ (or named paths) via Hermes/Mac; for each matched git remote, run check then sync.
  • Local execution = Hermes. SWE orchestrates (names repo, goal, whether the night uses Cloud vs Hermes/OmniRouter).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.