kf-update

name	kf-update
description	Update Kiloforge skill definitions and CLI tools from the latest release

Kiloforge Update

Update skill definitions in ~/.claude/skills/ and CLI tools in ~/.kf/bin/ from the latest GitHub release.

Use this skill when

• You want to update skills and CLI tools to the latest release
• A new skill or tool has been added and you need it
• A bug fix was made and you need the fix

Do not use this skill when

• The project has no .agent/kf/ directory (use /kf-setup first)
• You need to change project configuration (edit the yaml files directly)

Instructions

Step 1 — Verify Kiloforge is initialized

Check that ~/.kf/bin/ exists:

ls ~/.kf/bin/*.py

If not found, suggest /kf-setup instead. HALT.

Step 2 — Check current and latest versions

Check the installed version:

CURRENT_VERSION=""
if [ -f ~/.kf/VERSION ]; then
  CURRENT_VERSION=$(cat ~/.kf/VERSION)
fi
echo "Installed: ${CURRENT_VERSION:-unknown}"

Check the latest release version from GitHub:

LATEST_TAG=$(gh release view --repo Kiloforge/kiloforge-skills --json tagName -q '.tagName' 2>/dev/null)
if [ -z "$LATEST_TAG" ]; then
  echo "ERROR: Could not fetch latest release. Check network and gh auth."
  # HALT
fi
LATEST_VERSION="${LATEST_TAG#v}"
echo "Latest release: $LATEST_VERSION ($LATEST_TAG)"

If CURRENT_VERSION equals LATEST_VERSION, report that skills are already up to date and HALT (unless the user explicitly wants to force update).

Step 3 — Fetch the latest release

Clone the release tag to a temporary directory:

KF_TMPDIR=$(mktemp -d)
git clone --depth 1 --branch "$LATEST_TAG" \
  https://github.com/Kiloforge/kiloforge-skills.git \
  "$KF_TMPDIR/kiloforge-skills"

If the clone fails, HALT — the update cannot proceed without the release source.

Step 4 — Run the install script in update mode

python3 "$KF_TMPDIR/kiloforge-skills/kf-bin/scripts/kf-install.py" --update

This replaces:

• Skill definitions in ~/.claude/skills/
• CLI scripts in ~/.kf/bin/
• Writes ~/.kf/VERSION with the installed version

No per-project changes are made in update mode.

Step 4b — Clean up

rm -rf "$KF_TMPDIR"

Step 5 — Run per-project migrations (if needed)

If the current project has legacy per-project bin files at .agent/kf/bin/, clean them up:

if [ -d .agent/kf/bin ]; then
  echo "Cleaning legacy per-project bin directory..."
  rm -rf .agent/kf/bin
  git add -A .agent/kf/bin
  git diff --cached --quiet || git commit -m "chore(kf): remove legacy per-project bin/ (now global at ~/.kf/)"
fi

If running from a worktree (not the primary branch), merge using the standard protocol:

CURRENT_BRANCH=$(git branch --show-current)
PRIMARY_BRANCH=$(~/.kf/bin/kf-primary-branch.py 2>/dev/null || echo "main")
if [ "$CURRENT_BRANCH" != "$PRIMARY_BRANCH" ]; then
  ~/.kf/bin/kf-merge.py --holder "$(basename $(pwd))" --timeout 0
fi

This is a metadata-only merge (no --verify needed). If exit code 2 (lock held), report and retry. If exit code 3 (conflicts), resolve while locked and re-run.

Step 6 — Report

Show the output from kf-install.py — it reports which skills were added/updated and which scripts were copied.

Report the version change:

Updated: ${CURRENT_VERSION:-unknown} -> $LATEST_VERSION ($LATEST_TAG)