diff --git a/.claude/skills/wait/SKILL.md b/.claude/skills/wait/SKILL.md
new file mode 100644
index 0000000..09cd906
--- /dev/null
+++ b/.claude/skills/wait/SKILL.md
@@ -0,0 +1,25 @@
+---
+name: wait
+description: Pause execution for a requested number of minutes by sleeping one-minute increments to avoid exceeding shell timeouts.
+user_invocable: true
+triggers:
+  - /wait
+  - wait
+---
+
+# Wait Skill
+
+Use this skill whenever the user asks the assistant to pause or wait for a few minutes during a task. Instead of issuing a single long `sleep` command (which often hits the 2-minute shell timeout), run one-minute sleeps repeatedly for the requested duration.
+
+## Workflow
+
+1. **Determine the wait time.** Parse the user’s request for a duration expressed in minutes. If the request is vague, ask a clarifying question (e.g., “How many minutes should I wait?”) before running commands.
+2. **Enforce sane limits.** If the user requests a very large number of minutes, warn them and offer to break the wait into smaller chunks or confirm before proceeding.
+3. **Execute sequential Bash sleeps.** For each of the requested N minutes, issue a separate `bash` tool call with `sleep 60`. Before each call, report the upcoming iteration as `executing sleep: i/N: bash sleep 60` so observers know how many sleeps will run. Avoid bundling the sleeps into a single script; the goal is to keep every sleeping command under the 2-minute timeout.
+
+4. **Report completion.** Once the loop finishes, notify the user that the wait is over and resume the primary task.
+
+## Error handling
+
+- If the shell command fails (e.g., `sleep` unavailable), report the failure and stop waiting.
+- If the user changes their mind mid-wait, cancel the remaining iterations and explain how much time was actually spent waiting.
diff --git a/.github/workflows/linter_require_ruff.yaml b/.github/workflows/linter_require_ruff.yaml
index 5e6b4a4..3d75a39 100644
--- a/.github/workflows/linter_require_ruff.yaml
+++ b/.github/workflows/linter_require_ruff.yaml
@@ -34,5 +34,7 @@ jobs:
         run: make tech_debt
       - name: Run duplicate code check
         run: make duplicate_code
+      - name: Run AI writing check
+        run: uv run python scripts/check_ai_writing.py
       - name: Run import-linter
         run: make import_lint
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 9a4de1a..86c8cbd 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -41,3 +41,9 @@ repos:
         language: system
         pass_filenames: false
         always_run: true
+      - id: ai-writing-check
+        name: AI writing check
+        entry: uv run python scripts/check_ai_writing.py
+        language: system
+        pass_filenames: false
+        always_run: true
diff --git a/CLAUDE.md b/CLAUDE.md
index b814590..b7aa3a7 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -137,6 +137,7 @@ Structure as: `init()` → `continue(id)` → `cleanup(id)`
 ## Git Workflow
 - **Protected Branch**: `main` is protected. Do not push directly to `main`. Use PRs.
 - **Merge Strategy**: Squash and merge.
+- **Never force push**: Do not use `git push --force` or `--force-with-lease`. If you hit a git issue, stop and ask the user for guidance.
 
 ## Deprecated
 
diff --git a/scripts/check_ai_writing.py b/scripts/check_ai_writing.py
new file mode 100644
index 0000000..184e5b1
--- /dev/null
+++ b/scripts/check_ai_writing.py
@@ -0,0 +1,92 @@
+from __future__ import annotations
+
+import pathlib
+from collections.abc import Iterable, Sequence
+
+REPO_ROOT = pathlib.Path(__file__).resolve().parent.parent
+EM_DASH = chr(0x2014)
+SKIP_DIRS = {
+    ".git",
+    ".venv",
+    ".uv_cache",
+    ".uv-cache",
+    ".uv_tools",
+    ".uv-tools",
+    ".cache",
+    ".pytest_cache",
+    "__pycache__",
+    "node_modules",
+    ".next",
+}
+SKIP_SUFFIXES = {
+    ".png",
+    ".jpg",
+    ".jpeg",
+    ".gif",
+    ".webp",
+    ".ico",
+    ".mp4",
+    ".mov",
+    ".mp3",
+    ".woff",
+    ".woff2",
+    ".ttf",
+    ".otf",
+    ".eot",
+    ".pdf",
+    ".zip",
+    ".tar",
+    ".gz",
+    ".bz2",
+    ".7z",
+    ".ckpt",
+    ".bin",
+    ".pyc",
+    ".pyo",
+    ".db",
+}
+
+
+def iter_text_files(root: pathlib.Path) -> Iterable[pathlib.Path]:
+    for path in root.rglob("*"):
+        if not path.is_file():
+            continue
+        rel = path.relative_to(root)
+        if any(part in SKIP_DIRS for part in rel.parts):
+            continue
+        if path.suffix.lower() in SKIP_SUFFIXES:
+            continue
+        yield path
+
+
+def find_em_dashes(path: pathlib.Path) -> Sequence[tuple[int, str]]:
+    try:
+        text = path.read_text(encoding="utf-8", errors="ignore")
+    except OSError:
+        return []
+    lines: list[tuple[int, str]] = []
+    for lineno, line in enumerate(text.splitlines(), start=1):
+        if EM_DASH in line:
+            lines.append((lineno, line))
+    return lines
+
+
+def main() -> int:
+    violations: list[tuple[pathlib.Path, int, str]] = []
+    for path in iter_text_files(REPO_ROOT):
+        for lineno, line in find_em_dashes(path):
+            violations.append((path.relative_to(REPO_ROOT), lineno, line.strip()))
+    if violations:
+        print(
+            f"AI writing check failed: {EM_DASH!r} (em dash) detected in the repository"
+        )
+        for rel_path, lineno, snippet in violations:
+            print(f"{rel_path}:{lineno}: {snippet}")
+        print("Please remove the em dash or explain why it is acceptable.")
+        return 1
+    print("AI writing check passed (no em dash found).")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())