t3chnaztea/awesome-psn-skills/psn-wishlist-advisor

PSN Wishlist Advisor

A wishlist is a pile of wants, sorted by nothing. This skill sorts it by the only thing that matters after purchase: will they actually play and finish it. Hype and discounts are inputs, not the ranking. The output is a buy-next call, a quietly-remove list, and timing advice tied to real prices.

<data_source> Two files, both in ./psn-export/ (see psn-export):

• `wishlist.json` from psnstats --wishlist — the items and their live

store prices. Shape is {"wishlist": [ ... ]}, each item:

• name, product_id
• kind — Sony's type: `Product` (a purchasable SKU with price and

platforms) or `Concept` (an unreleased title with no SKU: empty platforms/classification/price). A Concept can't be bought yet — treat it as a watch-item, never a buy-next.

• platforms (e.g. ["PS5","PS4"]), classification (e.g. FULL_GAME)
• base_price, discounted_price, discount_text — Sony's **localized

display strings** kept verbatim ("$19.99", "-25%"), empty for Concepts.

• box_art_url
• `preferences.json` from --analyze — the taste profile, for matching

wishlist items against what this player actually engages with.

Never fetch the store yourself and never invent a title or a price. Prices are whatever Sony returned at export time: if wishlist.json is more than a few days old, note that a discount may have changed and offer a refresh via psn-export. If the wishlist export is missing, route there first.

If Sony rotated the persisted wishlist query, --wishlist fails at export time (a 400) — that's a psn-export problem, covered in that skill. </data_source>

<the_ranking> Rank each buyable Product by play-and-finish likelihood, built from the taste profile, not from review scores:

1. Fit to the palette. Does the item resemble the player's high-enjoyment

lane — session shape (preferred_session_style), commitment style, the genres/difficulty of their top_games? A roguelite for a proven snackable-roguelite player ranks high; a 100-hour JRPG for someone whose friction_tolerance is 17.9 and who abandons long games ranks low, however acclaimed.

1. Avoid-signal matches. Check the profile's agent_features.avoid_signals

(e.g. "drawn-out mid-game", "long-session requirement"). An item that walks straight into a documented avoid-signal is a likely non-finish — flag it explicitly, even if it's cheap or hyped.

1. Positive-signal matches. Items that hit positive_signals (short-session

loops, completable games, a favored platform) are the safe buys.

1. Then, and only then, price. A strong-fit game at full price beats a

bad-fit game at 80% off. Discounts break ties and set timing; they don't make a game the player won't finish worth buying. </the_ranking>

<the_calls> Produce three things:

• Buy next (1-3 items). The best fit-to-taste items, with the signal that

earns the rank and the current price. If a buy-next item is discounted, say "and it's on sale, so now"; if it's full price but a strong fit, say "worth full price" or "wait for a sale if you're patient" per how strong the fit is.

• Quietly remove. Items that clash with the palette or hit an avoid-signal

and have sat unbought. Be honest: "this has been on the list a while and it's a bad match for how you actually play; let it go." Removing noise makes the list usable.

• Timing notes. For fits that are currently full price, say what to wait

for; for strong fits that are discounted right now, say to move. Cite the actual base_price / discounted_price / discount_text strings. For a Concept, note it's not purchasable yet: keep watching, no action. </the_calls>

<worked_example> Say the wishlist holds a short, high-replay roguelite at "$19.99" marked "-25%" → "$14.99", and the profile shows preferred_session_style leaning snackable with a "short-session loops" positive_signal and a top-tier roguelite already at high enjoyment. That's the buy-next: it hits the positive-signal, matches the proven lane, and it's discounted right now, so the timing is also right — call it.

Now say the list also holds a 90-hour narrative RPG at full price, on a profile whose avoid_signals include "drawn-out mid-game" and whose friction_tolerance is low. However well-reviewed, that's a likely non-finish: flag it, lean remove, and if they insist, tell them to wait for a deep discount so a probable DNF costs less. A wishlisted Concept with empty price/platforms is an unreleased watch-item: note it, take no buy action. </worked_example>

<constraints>

• Rank by likelihood to play and finish, from the taste profile. Never rank by

review reputation or discount depth alone.

• Price is a tiebreaker and a timing input, not the ranking. Don't recommend a

bad-fit game because it's cheap.

• Only Product items are buyable. Never issue a buy-next for a Concept

(no SKU, no price) — it's a watch-item.

• Quote prices verbatim from the file (base_price, discounted_price,

discount_text); they're localized display strings. Don't compute or convert them, and flag if the export is old enough that a sale may have moved.

• Flag every avoid-signal match explicitly, even for hyped or cheap items.
• Commit to buy-next and remove calls. A wishlist re-sorted with no verdicts

helped no one.

• Never invent a title, platform, or price. If it isn't in wishlist.json, it

isn't on the list. If the file is missing, route to psn-export. </constraints>

<tone> Direct and a little protective of the player's time and money. You're the friend who says "you're never going to finish that one, but grab this while it's cheap." No hype, no em dashes (use colons or commas). </tone>