Fix pre-existing CI failures from the MCP-server commit

CI on main was already red before this branch. Address the repo-wide
failures that `pre-commit run --all-files` and the test job surface:

- requirements.txt: add `mcp>=1.9` so the CI/dev env can import
  steam_backlog_enforcer._mcp and collect test_mcp.py (mcp stays an
  optional dep in pyproject for the light systemd/CLI path).
- _stats.py: hoist the SteamAPIError import to module scope (PLC0415)
  and wrap an over-length f-string (E501).
- _web_dataset.py, docs/pick_game_manually_design.md: codespell typo
  fixes (unparsable, achievements, aggressively, until) + trailing
  newline.
- scripts/setup_mcp.sh: split declare/assign for SCRIPT_DIR/REPO_DIR
  (shellcheck SC2155).

Tests remain 100% covered.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RRCdR4D8SaVrN2xraMQxM

requirements.txt

@@ -5,6 +5,7 @@ bandit>=1.7.0
 codespell>=2.2.0
 coverage>=7.4.0
 howlongtobeatpy>=1.0
+mcp>=1.9
 mypy>=1.8.0
 pre-commit>=3.6.0
 pylint>=3.0.0

scripts/setup_mcp.sh

@@ -15,8 +15,10 @@
 
 set -euo pipefail
 
-readonly SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-readonly REPO_DIR="$(cd "${SCRIPT_DIR}/.." && pwd)"
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+readonly SCRIPT_DIR
+REPO_DIR="$(cd "${SCRIPT_DIR}/.." && pwd)"
+readonly REPO_DIR
 readonly VENV_DIR="${HOME}/.venvs/steam-backlog-mcp"
 
 main() {

steam_backlog_enforcer/_stats.py

@@ -35,7 +35,11 @@ from steam_backlog_enforcer.protondb import (
     ProtonDBRating,
     fetch_protondb_ratings,
 )
-from steam_backlog_enforcer.steam_api import GameInfo, SteamAPIClient
+from steam_backlog_enforcer.steam_api import (
+    GameInfo,
+    SteamAPIClient,
+    SteamAPIError,
+)
 
 if TYPE_CHECKING:
     from steam_backlog_enforcer.config import Config, State
@@ -276,7 +280,8 @@ def _print_pace_scenario(state: State, remaining: int, games_done: int) -> None:
     rate = games_done / days_elapsed
     _echo(f"    Started:        {started.strftime('%Y-%m-%d')}")
     _echo(
-        f"    Finished:       {games_done} games in {days_elapsed} days (since enforcement start)"
+        f"    Finished:       {games_done} games in {days_elapsed} days "
+        "(since enforcement start)"
     )
     _echo(
         f"    Pace:           {rate:.4f} games/day  (1 game every {1 / rate:.1f} days)"
@@ -357,8 +362,6 @@ def _refresh_recently_played_completions(
     except OSError:
         return games
 
-    from steam_backlog_enforcer.steam_api import SteamAPIError
-
     try:
         client = SteamAPIClient(config.steam_api_key, config.steam_id)
         owned_raw = client.get_owned_games()

steam_backlog_enforcer/_web_dataset.py

@@ -246,7 +246,7 @@ def count_complete_since_start(games: list[GameInfo], started_at: str) -> int:
     Games with no achievement timestamp data are excluded — their completion
     date is unknown, and they were most likely finished before Steam began
     recording unlock timestamps (i.e. before the enforcement period).
-    Returns 0 when started_at is empty or unparseable.
+    Returns 0 when started_at is empty or unparsable.
     """
     if not started_at:
         return 0

steam_backlog_enforcer/docs/pick_game_manually_design.md

@@ -3,11 +3,11 @@ Picking a game should work like this:
 1. user invokes script with specific flag for picking a game manually
 2. user provides game steam id (in testing use 489830)
 3. Script shows what game it believes this id means (in this case it should show The Elder Scrolls V: Skyrim Special Edition)
-4. user confirms that this is the game they want to pick and confirm that they will not be able to use the script for up to 2 weeks or until the game is completed (100% achievments)
+4. user confirms that this is the game they want to pick and confirm that they will not be able to use the script for up to 2 weeks or until the game is completed (100% achievements)
 
 When user picks a game manually this should override the current pick if it exists
-After picking manually backlog enforcer should make a note of that and very aggresively disallow user to do anything else
-for a period of 2 weeks or untill user completes a given game
+After picking manually backlog enforcer should make a note of that and very aggressively disallow user to do anything else
+for a period of 2 weeks or until user completes a given game
 Logic should be as follows:
     1. backlog checks if a user picked game manually
         a. if not -> continue as before