github-sync-helper

Objectives

1. Clear command quick reference (explanation + when to use)
2. Executable one-click scripts (avoid repeated manual steps)
3. Add numbers to common outputs as much as possible to facilitate your direct reply with number selection (e.g., repository list)

Security and Constraints (Must Follow)

$GITHUB_TOKEN

git status

Execution Method

• Script Entry:

  sh /var/minis/skills/github-sync-helper/scripts/gh_sync.sh <command> [options]

The script runs in the "current git repository directory" (must be inside the repository).

Git/GitHub Basic Operations Quick Reference (Explanation + Corresponding Script)

git init

init

git clone <url>

clone

git remote -v/add/set-url/remove

remotes/add-remote/add-upstream/set-remote-url/remove-remote

git branch -a/-d/-D

branches/create-branch/delete-branches

git switch <b>

checkout

git status

git diff

git log

status/diff/log

git add -A

add

git commit -m "..."

commit

git fetch

git pull

git push

fetch/pull/push/push-main

git merge

git rebase

merge/rebase

git stash

stash

git tag

tag

git submodule

submodule

gh-issues-list

Note: The script is positioned to "turn common basic operations into reusable commands". For complex rebase/conflicts, it is still recommended to use interactive terminal processing.

GitHub Platform Operations (API)

Unified requirement: Requires env

GITHUB_TOKEN

.

gh-issue-create --repo <owner/repo> --title <t> [--body <b>]

gh-issue-close --repo <owner/repo> --number <n>

gh-labels-list --repo <owner/repo>

gh-label-create --repo <owner/repo> --name <n> [--color <rrggbb>] [--description <d>]

gh-milestone-create --repo <owner/repo> --title <t> [--description <d>] [--due <YYYY-MM-DD>]

gh-releases-list --repo <owner/repo>

gh-actions-list --repo <owner/repo>

gh-actions-dispatch --repo <owner/repo> --workflow <id_or_file> [--ref <branch>] [--inputs <json>]

Script Command List (gh_sync.sh)

clone --url <url> [--dir <path>]

remotes

add-remote --name <n> --url <url>

set-remote-url --name <n> --url <url>

remove-remote --name <n>

add-upstream --upstream <owner/repo>

status

git status --porcelain

diff [--staged]

log [--n <k>]

branches

create-branch --name <b> [--from <ref>]

checkout --name <b>

delete-branches --keep <branch>

add --path <p>

git add

commit --message <m>

git commit

fetch [--remote <n>]

pull [--remote <n>] [--branch <b>]

push [--remote <n>] [--branch <b>]

push-main

empty-dir --dir <path>

restore-dir --src <path> --dst <path>

pr --upstream <owner/repo> --head <owner:branch> --base <branch> --title <t> --body <b>

gh-issues-list ...

Typical Workflows (Examples)

1) Push directly to main: Empty directory → Restore directory → Push (the set you just used)

sh /var/minis/skills/github-sync-helper/scripts/gh_sync.sh empty-dir --dir self-improving-agent
sh /var/minis/skills/github-sync-helper/scripts/gh_sync.sh restore-dir --src /var/minis/skills/self-improving-agent --dst self-improving-agent
sh /var/minis/skills/github-sync-helper/scripts/gh_sync.sh commit --message "restore(self-improving-agent): sync from local"
sh /var/minis/skills/github-sync-helper/scripts/gh_sync.sh push-main

4) Only replace the content of target files in the repository (retain original path/file name), then commit and push

# 1. Pull/refresh the repository
repo_dir=/var/minis/workspace/<repo>
if [ -d "$repo_dir/.git" ]; then
  cd "$repo_dir" && git fetch --all --prune && git reset --hard origin/main
else
  gh repo clone <owner/repo> "$repo_dir"
fi

# 2. Only content overwrite: Overwrite the target file in the repository with the source file content
cp <source_file> "$repo_dir/<target_path>"

# 3. If Git identity is not configured, prioritize reusing GitHub login name and use GitHub noreply email
cd "$repo_dir"
git config user.name '<GitHub Display Name>'
git config user.email '<login>@users.noreply.github.com'

# 4. Commit and push directly to main (only execute when the user explicitly requests commit/push)
git add <target_path>
git commit -m 'replace <target_path> content'
git push origin main

• When the user says "only content replacement", it is uniformly understood as: Retain the original file path and file name in the repository, only overwrite the content.
• Do not directly put the source file name into the repository; the target should still be the original file in the repository (e.g.,

  worker.js

  ).
• If the target path is known, directly overwrite the path; if unknown, locate the target file in the repository first before replacement.
• If

  Author identity unknown

  is reported before commit, write the following in the current repository:

  • git config user.name '<GitHub Display Name>'

  • git config user.email '<login>@users.noreply.github.com'

• By default, run

  git fetch

  +

  git reset --hard origin/main

  first to avoid accidental commits on the old working tree.
• If the user has explicitly requested "commit" or "push", you can continue execution directly without repeated confirmation.

5) Delete all branches except main (local + remote)

sh /var/minis/skills/github-sync-helper/scripts/gh_sync.sh delete-branches --keep main

6) No branch creation, directly open PR from fork:main → upstream:main

sh /var/minis/skills/github-sync-helper/scripts/gh_sync.sh pr \
  --upstream OpenMinis/MinisSkills \
  --head mowenyun:main \
  --base main \
  --title "sync: ..." \
  --body "..."

• Repository list output must use Markdown table and include a

  No.

  column to facilitate users to directly reply with numbers for selection.
• Recommended fields:

  No. | Repository(owner/repo) | Visibility | Fork | Default Branch | Last Updated | URL

• Interaction suggestion: After the table, prompt "Reply with the number to continue (clone/pull/commit and push/only content replacement, etc.)".

Note: If users need more fields (description, language, stars), add

--json ...

to expand.

could not read Username

GITHUB_TOKEN

GIT_ASKPASS

not inside a git repo

cd

clone

init