feat(kvb-coach): coach the operator's next step and remember what they enjoy

kvb-coach reads the situation (local time, hours at work today, the profile, open operator commands from kvb-org-operator, delegable kvb-org-rasci tasks) and suggests at most one next step per area: psyche, body, sleep, food, the operator's own tasks, and tasks to delegate to agents; short, friendly, with emojis, ordered by what the situation needs and biased to what the operator likes. It runs kvb-think-listen often to learn preferences. coach.py keeps the host-local profile in the gitignored .state/ of the skills repository: likes and dislikes per area (one replaces the other), a capped done-log, and the hours since the last activity per area; tests/test_coach.py covers it. General wellbeing only, no medical advice. Operator order 2026-11-03; two descriptions trimmed for the listing budget.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

README.md

@@ -9,7 +9,7 @@ Compatible with Claude Code, Codex, Gemini CLI, Cursor, Copilot, Windsurf, Cline
 
 ## Naming 🏷️
 
-Every first-party skill is named `kvb-<category>-<name>` with one of these categories: `think`, `create`, `code`, `ci`, `git`, `check`, `org`, `skill`. Only `kvb-cleanup`, `kvb-help` and the collectors `kvb-dream`, `kvb-optimize` and `kvb-transcend` stand without a category. Names use `a-z`, `0-9` and single hyphens, at most 64 characters, and equal their directory name. `tests/test_naming.py` enforces all of it, including that the organigram below lists every skill under its category.
+Every first-party skill is named `kvb-<category>-<name>` with one of these categories: `think`, `create`, `code`, `ci`, `git`, `check`, `org`, `skill`. Only `kvb-cleanup`, `kvb-coach`, `kvb-help` and the collectors `kvb-dream`, `kvb-optimize` and `kvb-transcend` stand without a category. Names use `a-z`, `0-9` and single hyphens, at most 64 characters, and equal their directory name. `tests/test_naming.py` enforces all of it, including that the organigram below lists every skill under its category.
 
 ## Organigram 🏛️
 
@@ -76,6 +76,7 @@ flowchart LR
   none --> transcend_all[kvb-transcend]
   none --> dream_all[kvb-dream]
   none --> optimize_all[kvb-optimize]
+  none --> coach_all[kvb-coach]
 ```
 
 ## Skill map 🗺️
@@ -158,6 +159,7 @@ flowchart LR
     transcend[kvb-transcend]
     dream[kvb-dream]
     optimize[kvb-optimize]
+    coach[kvb-coach]
   end
 
   commit_req -.-> commit_next
@@ -240,6 +242,9 @@ flowchart LR
   optimize --> refactor
   optimize --> tune
   optimize --> gaps
+  coach --> active_listening
+  coach --> operator
+  coach --> rasci
   autoskill --> autotune
   autotune --> essence
   transcend --> skill_transcend

skills/kvb-coach/SKILL.md

@@ -0,0 +1,73 @@
+---
+name: kvb-coach
+description: >
+  Coach the operator's next step: psyche, body, sleep, food, own and
+  delegable tasks. Trigger on /kvb-coach or "was soll ich tun".
+---
+
+Be the operator's coach: short, warm, concrete, with emojis, in the
+operator's language. Their health comes first; work is one area of six, not
+the default. Suggestions, never pressure.
+
+The profile keeps what the operator enjoys, dislikes and last did, host-local
+in the gitignored `.state/` of the skills repository:
+
+```bash
+coach.py --profile <skills repo>/.state/coach.json like|dislike|forget <area> "<thing>"
+coach.py --profile <same> done <area> "<thing>" [--at <iso>]
+coach.py --profile <same> show [--json]
+```
+
+Areas: `psyche`, `body`, `sleep`, `food`, `tasks`, `delegate`.
+
+## Step 1: read the situation
+
+- **Time**: local time and weekday, and how long the operator has been at it
+  today (their first message of the day in the session transcripts).
+- **Profile**: `coach.py show`: likes, dislikes, hours since the last sleep,
+  meal, workout or break.
+- **Own tasks**: the open commands of the `kvb-org-operator` skill and the
+  decisions agents wait on.
+- **Delegable**: what the operator did by hand that an agent could do, and
+  queued `kvb-org-rasci` register tasks without a worker.
+
+## Step 2: suggest
+
+At most one line per area, the most urgent first, the whole answer on one
+screen:
+
+- 🧠 **Psyche**: a short break, a walk outside, five minutes of meditation or
+  box breathing, a call with a friend, something just for fun.
+- 💪 **Body**: a workout from the operator's likes (a 20-minute run, a
+  bodyweight circuit, cycling, yoga), stretching neck, hips and back after
+  long sitting, standing up every hour.
+- 😴 **Sleep**: wind down, screens off, seven to nine hours; a 20-minute nap
+  in the early afternoon.
+- 🥗 **Food**: water, a real meal with vegetables and protein, a healthy
+  snack instead of sugar, no caffeine late in the day.
+- ✅ **Your tasks**: the one step only the operator can take that unblocks
+  the most.
+- 🤖 **Delegate**: what an agent can take over, with the skill or command
+  that does it.
+
+The situation sets the order: late evening puts sleep first, hours without
+a break put moving first, a meal more than five hours ago puts food first.
+Prefer what the operator likes, never offer a dislike, and do not repeat the
+last tip of an area. Close with one encouraging line.
+
+## Step 3: listen and remember
+
+Often, not every time: when the profile is empty, after a declined
+suggestion, and every few runs, run the `kvb-think-listen` skill with one
+short `AskUserQuestion` (what they enjoy, what fits right now, what to
+avoid). Record the answers with `like` and `dislike`, and what the operator
+reports having done with `done`.
+
+## Rules
+
+- The operator decides: no guilt, no nagging, a declined tip stays off for
+  the day.
+- General wellbeing, no medical advice: pain, illness or lasting exhaustion
+  get one kind pointer to a doctor.
+- The profile never leaves the host: not into commits, memory files or peer
+  messages.

skills/kvb-coach/scripts/coach.py

@@ -0,0 +1,135 @@
+#!/usr/bin/env python3
+"""Keep the operator's coaching profile: likes, dislikes and when what was done.
+
+Args:
+    --profile: path of the profile JSON file.
+    like|dislike <area> <thing>: record a preference; one replaces the other.
+    forget <area> <thing>: drop a preference.
+    done <area> <thing> [--at <iso>]: log that the operator did it.
+    show [--now <iso>] [--json]: likes, dislikes and the last logged activity
+        per area with the hours since.
+
+Exit codes:
+    0 done, 2 usage error.
+"""
+
+from __future__ import annotations
+
+import argparse
+import datetime
+import json
+import sys
+from pathlib import Path
+
+AREAS = ("psyche", "body", "sleep", "food", "tasks", "delegate")
+LOG_LIMIT = 500
+
+
+def fail(message: str) -> None:
+    print(f"coach: {message}", file=sys.stderr)
+    sys.exit(2)
+
+
+def moment(value: str | None) -> datetime.datetime:
+    if value is None:
+        return datetime.datetime.now(datetime.UTC)
+    try:
+        parsed = datetime.datetime.fromisoformat(value)
+    except ValueError:
+        fail(f"{value} is not an ISO timestamp")
+    return parsed if parsed.tzinfo else parsed.replace(tzinfo=datetime.UTC)
+
+
+def load(path: Path) -> dict:
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except FileNotFoundError:
+        return {"likes": {}, "dislikes": {}, "log": []}
+    except ValueError:
+        fail(f"{path} is not valid JSON")
+    return data
+
+
+def save(path: Path, data: dict) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    tmp = path.with_suffix(".tmp")
+    tmp.write_text(
+        json.dumps(data, indent=1, sort_keys=True, ensure_ascii=False), encoding="utf-8"
+    )
+    tmp.replace(path)
+
+
+def prefer(data: dict, keep: str, area: str, thing: str) -> None:
+    for key in ("likes", "dislikes"):
+        items = data[key].setdefault(area, [])
+        if thing in items:
+            items.remove(thing)
+    if keep:
+        data[keep][area].append(thing)
+
+
+def summary(data: dict, now: datetime.datetime) -> list[dict]:
+    rows = []
+    for area in AREAS:
+        last = next((e for e in reversed(data["log"]) if e["area"] == area), None)
+        hours = None
+        if last:
+            hours = round((now - moment(last["at"])).total_seconds() / 3600, 1)
+        rows.append(
+            {
+                "area": area,
+                "likes": data["likes"].get(area, []),
+                "dislikes": data["dislikes"].get(area, []),
+                "last": last["thing"] if last else None,
+                "hours_ago": hours,
+            }
+        )
+    return rows
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description=__doc__.splitlines()[0])
+    parser.add_argument("--profile", required=True)
+    sub = parser.add_subparsers(dest="cmd", required=True)
+    for name in ("like", "dislike", "forget"):
+        p = sub.add_parser(name)
+        p.add_argument("area", choices=AREAS)
+        p.add_argument("thing")
+    p = sub.add_parser("done")
+    p.add_argument("area", choices=AREAS)
+    p.add_argument("thing")
+    p.add_argument("--at")
+    p = sub.add_parser("show")
+    p.add_argument("--now")
+    p.add_argument("--json", action="store_true")
+    args = parser.parse_args()
+
+    path = Path(args.profile)
+    data = load(path)
+    if args.cmd == "show":
+        rows = summary(data, moment(args.now))
+        if args.json:
+            print(json.dumps(rows, indent=1, ensure_ascii=False))
+            return 0
+        for row in rows:
+            last = f"{row['last']} {row['hours_ago']}h ago" if row["last"] else "-"
+            print(
+                f"{row['area']}\tlast: {last}\tlikes: {', '.join(row['likes']) or '-'}"
+                f"\tdislikes: {', '.join(row['dislikes']) or '-'}"
+            )
+        return 0
+    if args.cmd == "done":
+        data["log"].append(
+            {"area": args.area, "thing": args.thing, "at": moment(args.at).isoformat()}
+        )
+        data["log"] = data["log"][-LOG_LIMIT:]
+    else:
+        keep = {"like": "likes", "dislike": "dislikes"}.get(args.cmd, "")
+        prefer(data, keep, args.area, args.thing)
+    save(path, data)
+    print(f"{args.cmd} {args.area} {args.thing}")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

skills/kvb-git-branch-gc/SKILL.md

@@ -1,10 +1,9 @@
 ---
 name: kvb-git-branch-gc
 description: >
-  Remove merged branches and worktrees with `pkgmgr branch gc`: dry run,
-  operator picks from the plan, then apply; unmerged items only one by one.
-  Trigger on /kvb-git-branch-gc, "branches aufräumen", "worktrees aufräumen",
-  "garbage collect".
+  Remove merged branches and worktrees with `pkgmgr branch gc` after a dry
+  run the operator picks from. Trigger on /kvb-git-branch-gc, "branches
+  aufräumen", "worktrees aufräumen".
 ---
 
 `pkgmgr branch gc` decides what is merged and does the removing; this skill

skills/kvb-org-matrix/SKILL.md

@@ -1,9 +1,8 @@
 ---
 name: kvb-org-matrix
 description: >
-  Print one matrix of everything in flight: tasks, workflows, background jobs,
-  with confidence and break risk per item. Trigger on /kvb-org-matrix,
-  "status", "übersicht", "wo stehen wir", or what is done, left or running.
+  One matrix of everything in flight with confidence and break risk per
+  item. Trigger on /kvb-org-matrix, "status", "übersicht", "wo stehen wir".
 ---
 
 One-shot overview, no side effects: this skill reads state and reports it. It

tests/test_coach.py

@@ -0,0 +1,66 @@
+"""Exercise coach.py: preferences flip between like and dislike, the log feeds hours since."""
+
+from __future__ import annotations
+
+import json
+import subprocess
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+
+REPO_ROOT = Path(__file__).resolve().parents[1]
+SCRIPT = REPO_ROOT / "skills" / "kvb-coach" / "scripts" / "coach.py"
+
+
+class TestCoach(unittest.TestCase):
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.addCleanup(self.tmp.cleanup)
+        self.profile = Path(self.tmp.name) / "state" / "coach.json"
+
+    def _run(self, *args: str) -> subprocess.CompletedProcess:
+        return subprocess.run(
+            [sys.executable, str(SCRIPT), "--profile", str(self.profile), *args],
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+
+    def _show(self, now: str) -> dict:
+        result = self._run("show", "--now", now, "--json")
+        self.assertEqual(result.returncode, 0, result.stderr)
+        return {row["area"]: row for row in json.loads(result.stdout)}
+
+    def test_a_dislike_replaces_a_like_and_forget_drops_it(self):
+        self._run("like", "body", "Laufen")
+        self._run("like", "body", "Laufen")
+        self._run("like", "body", "Yoga")
+        self._run("dislike", "body", "Laufen")
+        body = self._show("2026-11-03T18:00:00+00:00")["body"]
+        self.assertEqual(body["likes"], ["Yoga"])
+        self.assertEqual(body["dislikes"], ["Laufen"])
+        self._run("forget", "body", "Laufen")
+        self.assertEqual(self._show("2026-11-03T18:00:00+00:00")["body"]["dislikes"], [])
+
+    def test_the_latest_done_per_area_gives_the_hours_since(self):
+        self._run("done", "food", "Frühstück", "--at", "2026-11-03T08:00:00+00:00")
+        self._run("done", "food", "Mittagessen", "--at", "2026-11-03T12:30:00+00:00")
+        self._run("done", "sleep", "Nacht", "--at", "2026-11-02T23:00:00+00:00")
+        rows = self._show("2026-11-03T18:00:00+00:00")
+        self.assertEqual((rows["food"]["last"], rows["food"]["hours_ago"]), ("Mittagessen", 5.5))
+        self.assertEqual(rows["sleep"]["hours_ago"], 19.0)
+        self.assertIsNone(rows["psyche"]["last"])
+
+    def test_invalid_input_exits_2_and_keeps_the_profile(self):
+        self.assertEqual(self._run("like", "work", "x").returncode, 2)
+        self.assertEqual(self._run("done", "body", "x", "--at", "gestern").returncode, 2)
+        self.assertFalse(self.profile.exists())
+        self.profile.parent.mkdir(parents=True)
+        self.profile.write_text("{broken", encoding="utf-8")
+        self.assertEqual(self._run("show").returncode, 2)
+        self.assertEqual(self.profile.read_text(encoding="utf-8"), "{broken")
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_naming.py

@@ -11,7 +11,7 @@ SKILLS = REPO_ROOT / "skills"
 README = REPO_ROOT / "README.md"
 CATEGORIES = ("think", "create", "code", "ci", "git", "check", "org", "skill")
 UNCATEGORIZED = frozenset(
-    {"kvb-cleanup", "kvb-dream", "kvb-help", "kvb-optimize", "kvb-transcend"}
+    {"kvb-cleanup", "kvb-coach", "kvb-dream", "kvb-help", "kvb-optimize", "kvb-transcend"}
 )
 SPEC = re.compile(r"^[a-z0-9]+(-[a-z0-9]+)*$")
 SCHEMA = re.compile(rf"^kvb-(?:{'|'.join(CATEGORIES)})-[a-z0-9]+(?:-[a-z0-9]+)*$")