agentleFS
marcogallucciomarcogalluccio/company-brain-os/skills/ship/SKILL.md

ship

Skill7 starsChanged 48 days agoMITView on GitHub

Opens or updates the pull request for the active feature branch with the commits accumulated since main. The daily flow does not use this: daily work lands on main through /session-debrief. Part of the governance pack: only relevant once docs/GOVERNANCE.md step 2 (feature branches) is adopted, and it needs an authenticated gh CLI. Use when the operator says: ship, /ship, open the pr, raise the pr.

  • Commits and pushes

194 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ios-sync, open-gstack-browser and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Saqlain-811Skill

    ios-sync

    Regenerate the iOS debug bridge against the latest upstream gstack templates. (gstack)

    Also in 6 projects that use this

  • seek888Skill

    open-gstack-browser

    Launch GStack Browser — AI-controlled Chromium with the sidebar extension baked in. Opens a visible browser window where you can watch every action in real time. The sidebar shows a live activity feed and chat. Anti-bot stealth built in. Use when asked to "open gstack browser", "launch browser", "connect chrome", "open chrome", "real browser", "launch chrome", "side panel", or "control my browser". Voice triggers (speech-to-text aliases): "show me the browser".

    Also in 10 projects that use this

  • ayudaySkill

    setup-deploy

    Configure deployment settings for /land-and-deploy. Detects your deploy platform (Fly.io, Render, Vercel, Netlify, Heroku, GitHub Actions, custom), production URL, health check endpoints, and deploy status commands. Writes the configuration to CLAUDE.md so all future deploys are automatic. Use when: "setup deploy", "configure deployment", "set up land-and-deploy", "how do I deploy with gstack", "add deploy config".

    Also in 19 projects that use this

  • ShreyRaviSkill

    landing-report

    Read-only queue dashboard for workspace-aware ship. Shows which VERSION slots are currently claimed by open PRs, which sibling Conductor workspaces have WIP work likely to ship soon, and what slot /ship would pick next. No mutations — just a snapshot. Use when asked to "landing report", "what's in the queue", "show me open PRs", or "which version do I claim next". (gstack)

    Also in 9 projects that use this

  • zebbernSkill

    debug-fix

    Find and fix a bug or issue — from any source (GitHub issue, error message, user report, or observed behavior)

    Also in 4 projects that use this

  • ramybarsoumSkill

    document-release

    Post-ship documentation update. Reads all project docs, cross-references the diff, updates README/ARCHITECTURE/CONTRIBUTING/CLAUDE.md to match what shipped, polishes CHANGELOG voice, cleans up TODOS, and optionally bumps VERSION. Use when asked to "update the docs", "sync documentation", or "post-ship docs". Proactively suggest after a PR is merged or code is shipped.

    Also in 23 projects that use this

The file

We split this skill into the rules it gives an agent

12 rules. 0 are common, 5 are distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 180 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Without an argument, on main: do not stop empty-handed. The main tree stays on main in the daily flow, so the feature branch may well be …

    line 94
  2. Very specific:

    If there are uncommitted changes under daily-log/ or memory/: they never belong to the PR.

    line 127
  3. Very specific:

    If there are uncommitted changes outside daily-log/ and memory/: state plainly that they stay out of the PR too.

    line 134
  4. Very specific:

    If $COMMITS is empty, there is nothing to ship.

    line 146
  5. Distinctive:

    Check whether a PR is already open on this branch (the same check Step 3 runs, done early here to decide the stale-branch path):

    line 166
  6. Very specific:

    If $EXISTING is non-empty (a PR is already open): do not offer a rebase. Rewriting commits that are already pushed means the next push …

    line 173
  7. Very specific:

    When main truly has to enter the branch (GitHub reports conflicts, or the tests must rerun on the new main), the positive road is git merge …

    line 181
  8. Very specific:

    If non-empty, this run pushes to update that PR.

    line 204
  9. Distinctive:

    Ask once, with the commit count and the title:

    line 209
  10. Distinctive:

    Wait for confirmation, then run:

    line 214
  11. Very specific:

    If a PR already existed (Step 3), skip gh pr create and only run the git push -u: that alone updates the existing PR.

    line 223
  12. Distinctive:

    Never force-pushes: If the push is rejected because the remote diverged, stop and ask.

    line 247

“Common” is measured: kept by at least 6 in 10 of the 180 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

ship

Promote a feature branch's accumulated commits to a pull request. This is the other half of docs/GOVERNANCE.md step 2: structural work that earned a branch and a second pair of eyes gets closed out here, not through the debrief. It is on-demand, not tied to any session boundary; a feature branch can live for days and close with a single PR whenever it's ready.

Step 0: Resolve the operator and detect state

The operator-resolution block, identical to skills/session-debrief/SKILL.md:

cd "$(git rev-parse --show-toplevel)"
GIT_NAME="$(git config user.name)"
if [ -z "$GIT_NAME" ]; then
  echo "IDENTITY ERROR: git config user.name is empty. Set it as described in operators/ONBOARDING.md Step 4."
  exit 1
fi
MATCHES=""
for f in operators/*.md; do
  case "$f" in
    operators/CLAUDE.md|operators/ONBOARDING.md|operators/operator_template.md) continue ;;
  esac
  if awk '/^---$/{if(f)exit;f=1;next} f' "$f" | sed 's/^[[:space:]]*//' | grep -qxF -- "- $GIT_NAME"; then
    MATCHES="$MATCHES $f"
  fi
done
N=$(echo $MATCHES | wc -w | tr -d ' ')
if [ "$N" != "1" ]; then
  echo "IDENTITY ERROR: git user.name \"$GIT_NAME\" matches $N operator files:$MATCHES"
  echo "Exactly one operators/<slug>.md must claim this value in git_names."
  echo "Fix your identity or the registry (see operators/CLAUDE.md), then rerun."
  exit 1
fi
SLUG="$(basename $MATCHES .md)"
TODAY="$(date +%Y-%m-%d)"
echo "operator=$SLUG today=$TODAY"

On IDENTITY ERROR, stop the skill and show the message exactly as it printed. Never guess, never derive a slug from the git name itself: the registry rule in operators/CLAUDE.md is binding here too.

Then detect which branch this is:

CURRENT_BRANCH=$(git rev-parse --abbrev-ref HEAD)
echo "current=$CURRENT_BRANCH"

/ship takes an optional argument, the branch to ship (/ship <branch>). The argument wins over the current directory: when it is given, the branch is that one wherever the shell is, never the branch of the cwd.

With an argument. Find the worktree that has the branch checked out and work from there without asking: naming the branch was the choice.

TARGET_BRANCH="<the branch named in the invocation, as literal text>"
# porcelain prints "worktree <path>" two lines above "branch refs/heads/<name>";
# grep -Fx matches the branch line literally (a "." in the name is not a
# wildcard), then grep/cut pull the path rather than awk, so the snippet
# holds no positional parameters; cut -f2- keeps a path with spaces whole
WT_PATH=$(git worktree list --porcelain | grep -B2 -Fx "branch refs/heads/$TARGET_BRANCH" | grep "^worktree " | cut -d" " -f2-)
echo "worktree=${WT_PATH:-none}"
  • WT_PATH found: cd "$WT_PATH", recompute CURRENT_BRANCH, say "Shipping <branch> from <path>", and continue to Step 1. If $TARGET_BRANCH matches */proposals-* (a proposal pull request from skills/session-debrief/reference/split-paths.md Procedure B), stop instead: "<branch> is a proposals branch; it already gets its own PR from that flow. Use /pr-review to review it, not /ship."
  • The branch exists (git rev-parse --verify --quiet "refs/heads/$TARGET_BRANCH" succeeds) but no worktree has it: stop and ask. Shipping means checking the branch out somewhere, and the main tree stays on main; offer git worktree add <path> <branch> and rerun.
  • The branch does not exist: stop. "Branch <name> not found. Feature branches checked out in worktrees: <the git worktree list lines that are not main and not detached>."

Without an argument, on main: do not stop empty-handed. The main tree stays on main in the daily flow, so the feature branch may well be ready in a worktree. Run git worktree list. If it shows worktrees on feature branches (exclude the main tree, any detached one, and any branch matching */proposals-*, which belongs to the proposal pull request flow in skills/session-debrief/reference/split-paths.md Procedure B and already has its own PR), list them and ask which to ship; never move to one silently, the choice is the operator's when no argument was given:

"You're on main (the daily flow lands there through the debrief), but there are feature branches in worktrees: <branch, path>. Ship one of these? Name it, or rerun /ship <branch>."

On the answer, cd into the chosen worktree, recompute CURRENT_BRANCH, and continue to Step 1. If there is no such worktree, stop:

"You're on main: the daily flow already lands there through the debrief, and there is no feature branch in a worktree. /ship is only for feature branches (docs/GOVERNANCE.md step 2)."

Without an argument, on any other branch: confirm this is the branch to ship before doing anything else:

"You're on $CURRENT_BRANCH. Ship this branch as a PR?"

Wait for confirmation.

Step 1: Working tree check

git status --short

If there are uncommitted changes under daily-log/ or memory/: they never belong to the PR. Offer to run a debrief first: from a feature branch in a worktree, /session-debrief's branch guard commits only the branch's own work there and takes the daily log and memory to main from the main tree. If the operator would rather not, say plainly that those changes will not be part of the PR.

If there are uncommitted changes outside daily-log/ and memory/: state plainly that they stay out of the PR too. This skill never commits on the operator's behalf outside the debrief path above.

Step 2: Commits to ship

git fetch origin main --quiet
COMMITS=$(git log origin/main..HEAD --oneline)
BEHIND=$(git rev-list --count HEAD..origin/main)

If $COMMITS is empty, there is nothing to ship. Say so, and mention whether the branch is also behind ($BEHIND > 0 means it's stale but carries no commits of its own; $BEHIND = 0 means it's simply aligned with main). Exit either way.

Everything so far, Step 0's identity check, the branch detection, Step 1's working-tree check, and the fetch/count above, is plain git and never touches gh. From here on it does: the stale-branch path just below can call gh pr list, and Step 3 always does, so check the CLI is ready before going further. Run gh auth status. If it fails, say this skill needs an authenticated gh CLI, point at the dependency matrix in the README, and stop.

If $BEHIND > 0 (the branch is stale, forked from an older main): the two-dot commit list above (origin/main..HEAD) stays correct even from stale, it only ever shows the branch's own commits. The problem is the file-changed count that comes next: a two-dot diff --stat would also include everything main gained in the meantime, undone, which reads as a phantom payload far larger than what the branch actually carries.

Check whether a PR is already open on this branch (the same check Step 3 runs, done early here to decide the stale-branch path):

EXISTING=$(gh pr list --head "$CURRENT_BRANCH" --state open --json number,url -q '.[0]')
  • If $EXISTING is non-empty (a PR is already open): do not offer a rebase. Rewriting commits that are already pushed means the next push gets rejected as non-fast-forward, and this skill never force-pushes. Use git diff --stat origin/main...HEAD | tail -1 (three dots, merge-base) for the file-changed count only, never for the commit list. Say plainly: "The branch is behind main, but it already has an open PR, so I'm not rebasing (it would break the next push). The file-changed count below uses the three-dot diff so it doesn't include main's unrelated changes." When main truly has to enter the branch (GitHub reports conflicts, or the tests must rerun on the new main), the positive road is git merge origin/main on the branch: it adds one merge commit, rewrites no SHA, so the next push stays fast-forward, and the squash merge absorbs it at the end. A force-push, --force-with-lease included, is never the answer here without an explicit question in chat.
  • If $EXISTING is empty (no PR yet): offer a choice. "The branch is behind main by N commits. I can rebase onto origin/main first (needs a clean tree), or show the real payload with a three-dot diff without rebasing. Which do you want?" A rebase here needs the tree clean per Step 1; if it isn't, resolve that first. After a rebase, $BEHIND becomes 0 and the rest of this step proceeds on two dots as normal.

Otherwise ($BEHIND = 0), show the operator: the commit count and one-line subjects from $COMMITS, and the file-changed count from git diff --stat origin/main..HEAD | tail -1.

Step 3: Existing PR check

EXISTING=$(gh pr list --head "$CURRENT_BRANCH" --state open --json number,url -q '.[0]')

If non-empty, this run pushes to update that PR; it never opens a duplicate. Skip the gh pr create call in Step 4 and go straight to the push.

Step 4: Confirm, push, create

Ask once, with the commit count and the title:

"Open a PR with N commits onto main? Title: $TITLE (default: the branch name after the first /; tell me if you want something else)."

Wait for confirmation, then run:

git push -u origin "$CURRENT_BRANCH"
BODY_COMMITS=$(git log origin/main..HEAD --pretty=format:'- %s' --reverse)
gh pr create --title "$TITLE" --base main \
  --body "$(printf 'Feature branch %s into main (squash).\n\nCommits:\n%s\n' "$CURRENT_BRANCH" "$BODY_COMMITS")"

If a PR already existed (Step 3), skip gh pr create and only run the git push -u: that alone updates the existing PR.

On any push or create failure, show it plainly and stop. No silent retries.

Close with: "PR is up at <URL>. Run /pr-review <n> to review and merge it." Then, informational only, never run automatically: after the PR merges, git worktree remove <path> if the branch lived in one, then git branch -D <branch> to drop the local branch. /pr-review's merge gate already deletes the remote branch (--delete-branch); these two are local cleanup only, and only worth mentioning, never worth doing here. One expected message on that path: if the merge prints fatal: 'main' is already used by worktree, the merge on GitHub already succeeded and the error is about the local checkout of main only; verify with gh pr view <n> --json state (expect MERGED) rather than rerunning the merge.

Rules

  • One invocation, one PR action. Open a new PR or update an existing one, never both, never more than one PR per run.
  • Never pushes to main. This skill only ever pushes the feature branch itself.
  • Never auto-merges. Merging happens through /pr-review, so the operator sees the diff first.
  • Never force-pushes. If the push is rejected because the remote diverged, stop and ask; do not rebase and force past it. When main has to enter a branch with an open PR, merge it in (Step 2); a rebase of pushed commits is never offered.
  • Works from a worktree when the branch lives in one. An explicit /ship <branch> moves there without asking; on main without an argument, the skill lists the feature worktrees and asks, never moves silently.

Self-improvement

At the end of the run, check it against these friction triggers: avoidable round-trip · improvisation · breakage · token waste · operator correction (full definitions in skills/_improvements/capture.md). If at least one fired, ask the operator "there was friction on X, should I log it?"; on their ok, append an entry to skills/_improvements/friction-log.md in the format capture.md defines. Clean run: say nothing. Never self-edit this skill; improvements go through /skill-improve.

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.