@khoavu882/doflow 0.0.0-stage → 2.0.0-beta.4

package/core/shared/scripts/doflow/bash/do-git-state.sh

@@ -0,0 +1,546 @@
+#!/usr/bin/env bash
+# do-git-state.sh — deterministic lifecycle state helper for /do-git.
+#
+# Emits JSON describing the repository's position in the development cycle.
+# This is a deterministic layer component: no prompts, no side effects, just
+# reading git refs and producing structured data.
+
+set -uo pipefail
+
+if ! command -v jq >/dev/null 2>&1; then
+  printf '{"error":"jq-not-found","branch_class":null}\n'
+  exit 0
+fi
+
+script_dir="$(cd "$(dirname "$0")" && pwd)"
+resolved="$(bash "$script_dir/do-paths.sh" --json 2>/dev/null)"
+repo_root="$(printf '%s' "$resolved" | jq -r '.repo_root // empty')"
+# The resolver refuses a branch whose derived slug is unsafe and then prints no repo_root. This
+# helper does not need a feature, only the repository, so it asks git rather than reporting a
+# repository that exists as "not-a-git-repo".
+[ -z "$repo_root" ] && repo_root="$(git rev-parse --show-toplevel 2>/dev/null || true)"
+
+trunk_names="main develop"
+feature_prefixes="feat feature"
+fix_prefixes="fix bugfix"
+integration_branch="develop"
+production_branch="master"
+
+mode=""
+class_name=""
+slug=""
+
+for arg in "$@"; do
+  case "$arg" in
+    --state)                    mode="state" ;;
+    --next-version)             mode="next-version" ;;
+    --propagation-targets)      mode="propagation-targets" ;;
+    --fingerprint)              mode="fingerprint" ;;
+    --lifecycle)                mode="lifecycle" ;;
+    --branch-name)              mode="branch-name" ;;
+    --class=*)                  class_name="${arg#--class=}" ;;
+    --slug=*)                   slug="${arg#--slug=}" ;;
+    *) ;;
+  esac
+done
+
+[ -z "$mode" ] && mode="state"
+
+if [ -z "$repo_root" ]; then
+  printf '{"error":"not-a-git-repo"}\n'
+  exit 0
+fi
+
+cd "$repo_root"
+
+current_branch=""
+if git rev-parse --abbrev-ref HEAD >/dev/null 2>&1; then
+  current_branch="$(git rev-parse --abbrev-ref HEAD)"
+else
+  current_branch=""
+fi
+
+# ── Helper: classify a branch name ────────────────────────────────────────────
+get_class() {
+  local b="$1"
+  case "$b" in
+    "") echo "trunk" ;;
+    master|main|develop|HEAD) echo "trunk" ;;
+    release/*) echo "release" ;;
+    hotfix/*) echo "hotfix" ;;
+    feat/*|feature/*) echo "feature" ;;
+    fix/*|bugfix/*) echo "fix" ;;
+    */*) echo "other" ;;
+    *) echo "other" ;;
+  esac
+}
+
+get_slug() {
+  local b="$1"
+  case "$b" in
+    "") echo "" ;;
+    master|main|develop|HEAD) echo "" ;;
+    release/*|hotfix/*) echo "" ;;
+    */*) echo "${b#*/}" ;;
+    *) echo "$b" ;;
+  esac
+}
+
+commits_between() {
+  local from="$1"
+  local to="$2"
+  if git rev-parse "$from" >/dev/null 2>&1 && git rev-parse "$to" >/dev/null 2>&1; then
+    local ahead="$(git rev-list --count "${to}..${from}" 2>/dev/null || echo "0")"
+    local behind="$(git rev-list --count "${from}..${to}" 2>/dev/null || echo "0")"
+    echo "$ahead $behind"
+  else
+    echo "0 0"
+  fi
+}
+
+is_dirty() {
+  local status="$(git status --short 2>/dev/null | wc -l)"
+  [ "$status" -gt 0 ]
+}
+
+# Which ref the integration branch resolves to decides whether a distance measured against it
+# means anything. The local branch was used unconditionally, so a develop that had not been pulled
+# made every distance wrong by however stale it was — observed live at 22 reported against 32.
+#
+# The remote-tracking ref is preferred where it exists; the local branch is the fallback for a
+# repository with no remote or one never fetched. Preferring it does not make the answer *fresh* —
+# origin/develop is only as current as the last fetch — so the third field reports how far the
+# local branch trails, which is non-zero exactly when a `git fetch` would change the answer.
+#
+# Echoes: "<ref> <kind> <local-behind-remote>"
+resolve_integration_ref() {
+  local branch="$1"
+  local ref="$branch" kind="local" behind=0
+  if git rev-parse --verify --quiet "refs/remotes/origin/${branch}" >/dev/null 2>&1; then
+    ref="origin/${branch}"
+    kind="remote-tracking"
+    if git rev-parse --verify --quiet "refs/heads/${branch}" >/dev/null 2>&1; then
+      behind="$(git rev-list --count "${branch}..origin/${branch}" 2>/dev/null || echo 0)"
+    fi
+  fi
+  printf '%s %s %s\n' "$ref" "$kind" "$behind"
+}
+
+# Commits either side of a comparison, or "0 0" when either ref is missing.
+# Echoes: "<ahead> <behind>"
+distance_from() {
+  local branch="$1" against="$2"
+  if git rev-parse "refs/heads/${branch}" >/dev/null 2>&1 && \
+     git rev-parse --verify --quiet "$against" >/dev/null 2>&1; then
+    local counts; counts="$(commits_between "$branch" "$against")"
+    printf '%s %s\n' "$(echo "$counts" | awk '{print $1}')" "$(echo "$counts" | awk '{print $2}')"
+  else
+    printf '0 0\n'
+  fi
+}
+
+# The single lifecycle word for a branch's position. Ordered so the most actionable state wins:
+# an uncommitted change matters more than any distance, and being behind matters more than ahead.
+lifecycle_position() {
+  local ahead="$1" behind="$2" dirty="$3"
+  if [ "$dirty" = "true" ]; then printf 'dirty-worktree\n'
+  elif [ "$behind" -gt 0 ]; then printf 'behind-integration\n'
+  elif [ "$ahead" -gt 0 ]; then printf 'ahead-of-integration\n'
+  else printf 'in-sync\n'
+  fi
+}
+
+do_state() {
+  local class="$(get_class "$current_branch")"
+  local slugs="$(get_slug "$current_branch")"
+  local dirty="false"
+  is_dirty && dirty="true"
+  
+  local integration_ref integration_ref_kind integration_local_behind
+  read -r integration_ref integration_ref_kind integration_local_behind \
+    <<< "$(resolve_integration_ref "$integration_branch")"
+
+  local ahead behind position
+  read -r ahead behind <<< "$(distance_from "$current_branch" "$integration_ref")"
+  position="$(lifecycle_position "$ahead" "$behind" "$dirty")"
+  
+  jq -n \
+    --arg branch "${current_branch:-null}" \
+    --arg class_name "$class" \
+    --arg feature_slug "$slugs" \
+    --argjson dirty "$dirty" \
+    --arg ahead "$ahead" \
+    --arg behind "$behind" \
+    --arg position "$position" \
+    --arg integration_ref "$integration_ref" \
+    --arg integration_ref_kind "$integration_ref_kind" \
+    --arg integration_local_behind "$integration_local_behind" \
+    '{
+      branch:        (if $branch=="" then null else $branch end),
+      class:         $class_name,
+      feature_slug:  (if $feature_slug=="" then null else $feature_slug end),
+      dirty:         $dirty,
+      ahead_of_integration: ($ahead | tonumber),
+      behind_integration: ($behind | tonumber),
+      integration_ref: $integration_ref,
+      integration_ref_kind: $integration_ref_kind,
+      integration_local_behind_remote: ($integration_local_behind | tonumber),
+      position:      $position
+    }'
+}
+
+# The production branch's refs, for finding release tags that were never merged back: the local
+# branch and its remote-tracking ref, both when both exist, because either may be the one that
+# holds the newest release (a local `main` that was not pulled is stale, a remote-tracking one that
+# was not fetched is too). `main` is the default; `master` is the older default the lifecycle
+# policy still names, used only when no `main` ref exists. Echoes one ref per line, possibly none.
+production_refs() {
+  local b ref found
+  for b in main master; do
+    found=""
+    for ref in "refs/heads/${b}" "refs/remotes/origin/${b}"; do
+      if git rev-parse --verify --quiet "$ref" >/dev/null 2>&1; then printf '%s\n' "${ref#refs/}"; found=1; fi
+    done
+    [ -n "$found" ] && return
+  done
+}
+
+# Splits a semver string into its numeric core and its pre-release label.
+# Echoes: "<major> <minor> <patch> <prerelease-or-empty>"
+parse_semver() {
+  local version="$1"
+  local core="${version%%-*}"; core="${core%%+*}"
+  local pre=""
+  case "$version" in
+    *-*) pre="${version#*-}"; pre="${pre%%+*}" ;;
+  esac
+  local major minor patch
+  IFS='.' read -r major minor patch <<< "$core" 2>/dev/null || true
+  printf '%s %s %s %s\n' "${major:-0}" "${minor:-0}" "${patch:-0}" "$pre"
+}
+
+# Reads the commit subjects since a base tag and names the bump they imply.
+# Echoes: "<MAJOR|MINOR|PATCH|INITIAL> <commit-count>"
+classify_bump() {
+  local base_tag="$1"
+  if [ -z "$base_tag" ]; then
+    printf 'INITIAL 0\n'
+    return
+  fi
+  local subjects count kind="PATCH"
+  subjects="$(git log --format=%s "${base_tag}..HEAD" 2>/dev/null || true)"
+  # `git rev-list --count`, not `printf | wc -l`: the latter sees no trailing newline and so
+  # undercounts every range by one.
+  count="$(git rev-list --count "${base_tag}..HEAD" 2>/dev/null || echo 0)"
+  if printf '%s' "$subjects" | grep -qE 'BREAKING CHANGE|\!:'; then
+    kind="MAJOR"
+  elif printf '%s' "$subjects" | grep -qE '^feat|^feature'; then
+    kind="MINOR"
+  fi
+  printf '%s %s\n' "$kind" "$count"
+}
+
+# The next free number on a pre-release line. Computing label.N+1 arithmetically once proposed a
+# tag that already existed — a well-formed value derived without consulting the thing it describes.
+# `git describe` finds the nearest *reachable* tag, so a pre-release cut on an unmerged branch is
+# invisible to the base-tag lookup and collides here instead. Taken numbers are stepped over rather
+# than stopped on, because the caller wants a usable candidate, and the count of them is reported
+# because the gap is a fact about the tag history. Bounded so a pathological tag set cannot spin.
+# Echoes: "<candidate> <skipped-count>"
+next_free_prerelease() {
+  local core="$1" pre="$2"
+  local label="${pre%.*}" num="${pre##*.}"
+  if ! printf '%s' "$num" | grep -qE '^[0-9]+$'; then
+    printf '%s-%s.1 0\n' "$core" "$pre"
+    return
+  fi
+  local candidate=$((num + 1)) skipped=0 guard=0
+  while [ "$guard" -lt 100 ]; do
+    if ! git rev-parse --verify --quiet "refs/tags/v${core}-${label}.${candidate}" >/dev/null 2>&1; then
+      break
+    fi
+    candidate=$((candidate + 1)); skipped=$((skipped + 1)); guard=$((guard + 1))
+  done
+  printf '%s-%s.%s %s\n' "$core" "$label" "$candidate" "$skipped"
+}
+
+# The newest release tag reachable from HEAD or from the production branch, by version order;
+# empty when there is none. do_next_version bases its proposal on it and the lifecycle output
+# reports it, so the two never disagree. The reasoning behind "newest by version" and "reachable
+# from production too" is in do_next_version's comments below.
+find_base_tag() {
+  local base_tag="" eligible="" ref
+  eligible="$(git tag --merged HEAD --list 'v*' 2>/dev/null || true)"
+  while IFS= read -r ref; do
+    [ -z "$ref" ] && continue
+    eligible="${eligible}"$'\n'"$(git tag --merged "$ref" --list 'v*' 2>/dev/null || true)"
+  done <<< "$(production_refs)"
+  eligible="$(printf '%s\n' "$eligible" | sort -u | sed '/^$/d')"
+  if [ -n "$eligible" ]; then
+    base_tag="$(git tag --list 'v*' --sort=-v:refname 2>/dev/null | grep -Fx -f <(printf '%s\n' "$eligible") | head -1 || true)"
+  fi
+  printf '%s\n' "$base_tag"
+}
+
+do_next_version() {
+  # Both start set: `set -u` makes a bare `local base_tag` an unbound variable on bash 4+ whenever
+  # no eligible tag exists (a repository with no release tags yet).
+  local base_tag="" current_version="0.0.0"
+  # `git describe --tags --abbrev=0` answers "nearest reachable tag by commit distance", and every
+  # version decision below needs "newest reachable tag by version". Those coincide only while
+  # history is linear — and a release ritual merges twice, so the newest tag routinely sits further
+  # from HEAD than an older one. On this repository after v1.0.0 shipped, beta.8 was reachable at 33
+  # commits and beta.7 at 31, so describe returned beta.7 and every later computation was based on a
+  # superseded release.
+  #
+  # Sorting by version instead. The `v*` filter keeps a non-version tag from winning the sort, and
+  # --merged keeps the answer to tags that are actually released work.
+  #
+  # "Merged" means into HEAD *or into the production branch*. A release ritual tags the merge commit
+  # on production, and the integration branch only gets that commit back through a second merge
+  # that may not have happened yet, so from the integration branch the newest release tag is
+  # unreachable and the proposal was based on an older one (v1.8.3 proposed while v1.12.0 was out).
+  # The tags are sorted by git as before; the union only decides which of them are eligible.
+  base_tag="$(find_base_tag)"
+  [ -n "$base_tag" ] && current_version="${base_tag#v}"
+
+  # The manifest is what a release publishes. A base tag that disagrees with it means the tag or
+  # the manifest is behind, which the caller should see before proposing a version from either.
+  local manifest_version="" warning=""
+  if [ -n "$base_tag" ] && [ -f package.json ]; then
+    manifest_version="$(jq -r '.version // empty' package.json 2>/dev/null || true)"
+    if [ -n "$manifest_version" ] && [ "$manifest_version" != "$current_version" ]; then
+      warning="manifest version ${manifest_version} differs from base tag ${base_tag}"
+    fi
+  fi
+
+  local major minor patch prerelease
+  read -r major minor patch prerelease <<< "$(parse_semver "$current_version")"
+  local version_core="${major}.${minor}.${patch}"
+
+  local bump_kind commit_count
+  read -r bump_kind commit_count <<< "$(classify_bump "$base_tag")"
+
+  # Semver orders 1.0.0-beta.7 < 1.0.0 < 1.0.1, so a pre-release already sits *before* its own
+  # version number: promoting it IS the bump, not an increment on top of it. A bump promotes when
+  # the field it would raise is already zero, and otherwise increments with the suffix falling
+  # away. These are node-semver's inc() rules reproduced rather than invented.
+  local next_major="$major" next_minor="$minor" next_patch="$patch"
+  if [ "$bump_kind" = "INITIAL" ]; then
+    next_major=1; next_minor=0; next_patch=0
+  elif [ -n "$prerelease" ]; then
+    case "$bump_kind" in
+      MAJOR) [ "$minor" -eq 0 ] && [ "$patch" -eq 0 ] || { next_major=$((major + 1)); next_minor=0; next_patch=0; } ;;
+      MINOR) [ "$patch" -eq 0 ] || { next_minor=$((minor + 1)); next_patch=0; } ;;
+    esac
+  else
+    case "$bump_kind" in
+      MAJOR) next_major=$((major + 1)); next_minor=0; next_patch=0 ;;
+      MINOR) next_minor=$((minor + 1)); next_patch=0 ;;
+      *)     next_patch=$((patch + 1)) ;;
+    esac
+  fi
+  local next_version="${next_major}.${next_minor}.${next_patch}"
+
+  # The other honest option for a pre-release base: continue the line rather than promote it.
+  # Reported alongside so the release ritual can offer the choice instead of assuming one.
+  local next_prerelease="" next_prerelease_skipped=0
+  if [ -n "$prerelease" ]; then
+    read -r next_prerelease next_prerelease_skipped <<< "$(next_free_prerelease "$version_core" "$prerelease")"
+  fi
+
+  jq -n \
+    --arg base_tag "${base_tag:-null}" \
+    --arg current_version "$current_version" \
+    --arg next_version "$next_version" \
+    --arg next_prerelease "$next_prerelease" \
+    --arg next_prerelease_skipped "$next_prerelease_skipped" \
+    --arg prerelease "$prerelease" \
+    --arg bump_kind "$bump_kind" \
+    --argjson commit_count "$commit_count" \
+    --arg warning "$warning" \
+    '{
+      base_tag:         (if $base_tag=="" then null else $base_tag end),
+      current_version:  $current_version,
+      is_prerelease:    ($prerelease != ""),
+      next_version:     $next_version,
+      next_prerelease:  (if $next_prerelease=="" then null else $next_prerelease end),
+      next_prerelease_skipped: ($next_prerelease_skipped | tonumber),
+      bump_kind:        $bump_kind,
+      commits_count:    ($commit_count | tonumber)
+    } + (if $warning != "" then {warning: $warning} else {} end)'
+}
+
+do_propagation_targets() {
+  local targets_json="[]"
+  
+  # Integration branch is always a target
+  targets_json="$(echo "$targets_json" | jq --arg k "integration" --arg n "$integration_branch" --argjson hf "true" \
+    '. += [{kind: $k, name: $n, has_fix: $hf}]')"
+  
+  # Get all active release branches as targets
+  while IFS= read -r rel; do
+    [ -z "$rel" ] && continue
+    targets_json="$(echo "$targets_json" | jq --arg k "release" --arg n "$rel" --argjson hf "false" \
+      '. += [{kind: $k, name: $n, has_fix: $hf}]')"
+  done < <(git for-each-ref --format='%(refname:short)' refs/heads/release/* 2>/dev/null | sed 's|refs/heads/release/||')
+  
+  local current_class="$(get_class "$current_branch")"
+  
+  jq -n \
+    --arg current "${current_branch:-null}" \
+    --arg class "$current_class" \
+    '{current: (if $current=="" then null else $current end), class: $class, targets: []}' \
+    | jq --argjson t "$targets_json" '.targets = $t'
+}
+
+do_fingerprint() {
+  local head_sha=""
+  head_sha="$(git rev-parse HEAD 2>/dev/null | head -c 40)" || true
+  
+  [ -z "$head_sha" ] && { printf '{"error":"cannot-read-head-sha"}\n'; exit 1; }
+  
+  local ref_list=""
+  while IFS= read -r r; do
+    ref_list="${ref_list}${r}"$'\n'
+  done < <(git for-each-ref --format='%(refname)' refs/heads refs/tags 2>/dev/null | sort)
+  
+  local fp=""
+  fp="$(printf '%s\n%s' "$head_sha" "$ref_list" | shasum -a 256 | cut -d' ' -f1)" || true
+  
+  [ -z "$fp" ] && fp="fingerprint-unavailable"
+  
+  jq -n \
+    --arg sha "$head_sha" \
+    --arg branch "${current_branch:-null}" \
+    --arg fingerprint "$fp" \
+    '{
+      sha:    $sha,
+      branch: (if $branch=="" then null else $branch end),
+      fingerprint: $fingerprint
+    }'
+}
+
+do_branch_name() {
+  local class="${class_name:-}"
+  local slugs="${slug:-}"
+  
+  [ -z "$class" ] && { printf '{"error":"missing-class"}\n'; exit 2; }
+  [ -z "$slugs" ] && { printf '{"error":"missing-slug"}\n'; exit 2; }
+
+  # A slug becomes a branch name and then a directory under agent-docs/doflow/, so it must not be
+  # able to name a path or a flag (IC-009). Same shape do-paths.sh enforces for its own slugs.
+  # ASCII ranges whatever the locale: under a UTF-8 collation [A-Za-z] can admit "é".
+  if ! (export LC_ALL=C; [[ "$slugs" =~ ^[A-Za-z0-9][A-Za-z0-9._-]*$ ]] && [[ "$slugs" != *..* ]]); then
+    jq -n --arg slug "$slugs" '{error: "invalid-slug", message: ("slug \"" + $slug + "\" is not a valid feature slug: use letters, digits, dot, underscore or dash, start with a letter or digit, and no \"..\""), hint: "use letters, digits, dot, underscore or dash, start with a letter or digit, and no \"..\""}'
+    exit 2
+  fi
+  
+  # Two vocabularies meet here and only one of them is a branch prefix.
+  #
+  # The arms below are *branch* classes — the same set get_class emits — so --class=fix round-trips
+  # and --class=bug did not: it fell through to a catch-all that prepended whatever it was handed,
+  # producing bug/<slug>, which get_class then classified as "other". Every DoFlow skill holds a
+  # *task* class (bug, refactor, dependency-change, trivial-edit, feature), so the caller most
+  # likely to use this verb was the one guaranteed to get an unusable name.
+  #
+  # Task classes are mapped rather than passed through. In branch terms everything that is not a
+  # feature is a fix: the lifecycle policy declares exactly two working prefixes, feat and fix, and
+  # inventing refactor/ or trivial-edit/ would widen a vocabulary get_class does not recognise —
+  # trading one unclassifiable name for four.
+  local branch_class="$class"
+  case "$class" in
+    bug|refactor|dependency-change|trivial-edit) branch_class="fix" ;;
+  esac
+
+  local branch_name=""
+  case "$branch_class" in
+    feature)   branch_name="feat/${slugs}" ;;
+    fix)       branch_name="fix/${slugs}" ;;
+    release)   branch_name="release/${slugs}" ;;
+    hotfix)    branch_name="hotfix/${slugs}" ;;
+    trunk|other)
+      printf '{"error":"cannot-create-branch-for-trunk-or-other-class"}\n'
+      exit 2
+      ;;
+    # An unrecognised class is refused rather than prepended. The catch-all that used to sit here
+    # is what let a task class through and produced a name no verb could classify.
+    *)
+      printf '{"error":"unknown-class","class":"%s","valid":"feature, fix, release, hotfix, bug, refactor, dependency-change, trivial-edit"}\n' "$class"
+      exit 2
+      ;;
+  esac
+  
+  jq -n --arg name "$branch_name" '{name: $name}'
+}
+
+# The git facts the lifecycle verbs need, in one place so Node never derives them a second way
+# (feature 046, DEC-044, DEC-035). Read-only.
+#   integration_ref  develop, else main, else master, else origin/HEAD: the first that resolves, a
+#                    local branch before its origin/ tracking ref; null when none does. Unlike
+#                    resolve_integration_ref this never prefers the remote, because the lifecycle
+#                    views must agree with the release preview about one pinned ref.
+#   feature_prefixes the branch prefixes a feature branch is named with
+#   integration_local_behind_remote  how many commits origin/<integration> is ahead of the local
+#                    integration branch (0 when it has no origin counterpart or is itself origin/...);
+#                    the same field --state reports, so a merge that only exists on the remote is
+#                    visible as the staleness it is. No fetch is run: the answer is as fresh as the
+#                    last one.
+#   release_tags     v* tags that match the release-tag pattern (vX.Y.Z with an optional pre-release)
+#   base_tag         the same base tag --next-version proposes from, or null
+lifecycle_integration_ref() {
+  local name
+  for name in develop main master; do
+    if git rev-parse --verify --quiet "refs/heads/${name}" >/dev/null 2>&1; then printf '%s\n' "$name"; return; fi
+    if git rev-parse --verify --quiet "refs/remotes/origin/${name}" >/dev/null 2>&1; then printf 'origin/%s\n' "$name"; return; fi
+  done
+  if git rev-parse --verify --quiet "refs/remotes/origin/HEAD" >/dev/null 2>&1; then printf 'origin/HEAD\n'; return; fi
+  printf '\n'
+}
+
+do_lifecycle() {
+  local integration="" tags="" base_tag=""
+  integration="$(lifecycle_integration_ref)"
+  tags="$(git tag --list 'v*' 2>/dev/null | grep -E '^v[0-9]+\.[0-9]+\.[0-9]+(-[0-9A-Za-z.-]+)?$' || true)"
+  base_tag="$(find_base_tag)"
+
+  local behind=0
+  case "$integration" in
+    ""|origin/*) ;;
+    *)
+      if git rev-parse --verify --quiet "refs/remotes/origin/${integration}" >/dev/null 2>&1; then
+        behind="$(git rev-list --count "${integration}..origin/${integration}" 2>/dev/null || echo 0)"
+      fi
+      ;;
+  esac
+
+  local prefixes_json="" tags_json=""
+  prefixes_json="$(printf '%s\n' $feature_prefixes | jq -R . | jq -s .)"
+  tags_json="$(printf '%s' "$tags" | jq -R -s 'split("\n") | map(select(length > 0))')"
+
+  jq -n \
+    --arg integration "$integration" \
+    --arg base_tag "$base_tag" \
+    --arg behind "$behind" \
+    --argjson prefixes "$prefixes_json" \
+    --argjson tags "$tags_json" \
+    '{
+      integration_ref:  (if $integration=="" then null else $integration end),
+      integration_local_behind_remote: ($behind | tonumber),
+      feature_prefixes: $prefixes,
+      release_tags:     $tags,
+      base_tag:         (if $base_tag=="" then null else $base_tag end)
+    }'
+}
+
+case "$mode" in
+  state)              do_state ;;
+  next-version)       do_next_version ;;
+  propagation-targets) do_propagation_targets ;;
+  fingerprint)        do_fingerprint ;;
+  lifecycle)          do_lifecycle ;;
+  branch-name)        do_branch_name ;;
+  *)                  printf '{"error":"unknown-mode","mode":"%s"}\n' "$mode"; exit 2 ;;
+esac
+
+exit 0