← ClaudeAtlas

game-library-analysislisted

Pull a gaming library with playtime from Steam, Xbox, or PlayStation, build an Excel workbook analyzing it, and produce ranked "what should I play next" recommendations grounded in the actual hours rather than critical acclaim. Use this whenever the user mentions their Steam library, backlog, playtime, unplayed games, what to play next, "hidden gems", game recommendations based on what they already own, or wants to connect a gaming account so recommendations skip the games they have already played. Also use it when refreshing or extending an existing library workbook, or adding a second platform to one.
roloc/game-library-analysis · ★ 1 · Data & Documents · score 72
Install: claude install-skill roloc/game-library-analysis
# Game Library Analysis Turn a raw platform export into a workbook plus recommendations that are actually defensible — every pick traceable to hours the person really logged. The core insight this exists to serve: **people are unreliable narrators of their own taste.** They will tell you they love roguelikes when one outlier game accounts for 90% of the genre's hours, or that they "played that at launch" when the data says five hours and never again. The library is the ground truth. Lead with it, and be willing to contradict the person's self-description when the numbers disagree — that contradiction is usually the most valuable thing you produce. ## Workflow 1. Get the data — `python run.py` fetches every platform configured in `config.env`, merges, and builds the workbook + HTML report (see `references/platforms.md` for access routes and their tradeoffs) 2. Read `references/analysis.md` and do the analysis pass over the merged CSV 3. **Write `data/recommendations.json`** — the hand-written judgment (schema below). This step is yours, not a script's: verify release statuses and known issues by web search first, cite hours in every row. 4. Re-run `python run.py --no-fetch` — both the workbook and the report render the recommendations, skip list, on-hold list, and scoreboard. 5. Hand over both files and lead the writeup with the strongest finding. **Running the pipeline alone is not the job.** The charts are measurements; the recommendations are the point.