Cosa contiene

Un file SKILL.md (la guida operativa) + un modello di partenza. Un'anteprima:

AI-Use Disclosure Playbook

When to use

• Your studio uses AI tools and players, press, or platform holders are asking — or will ask.
• You're about to announce, ship, or patch something AI-assisted and want the disclosure to land as honesty, not confession.
• A team member (or executive) already said something blunt about AI in public and you need to manage the fallout.
• You're writing the studio's standing policy so the next disclosure isn't improvised.

Workflow

1. Map what you actually use (input: production reality → output: AI-use inventory)
   List every AI touchpoint honestly: code assistance, concept art exploration, localization drafts, QA bots, voice synthesis, marketing copy, community moderation. For each: which tool, at what stage, and — critically — what a human did after (reviewed, rewrote, approved, discarded). "AI-assisted with human approval" and "AI-generated, shipped as-is" are different disclosures; most backlash comes from blurring them.
2. Segment your audiences (input: inventory → output: audience map)
   Three audiences, three anxieties: players (fear: slop replacing craft, training-data ethics, job losses), press (fear: being misled; they will find the AI use anyway), platform holders/partners (fear: legal and content-policy exposure). Write one sentence per audience: what they need to hear to keep trusting you.
3. Write the disclosure (input: inventory + audience map → output: disclosure statement)
   Structure: what was used, where in the pipeline, what humans did, and what it changed for the player (faster patches? more localization languages? nothing visible?). Lead with the human work, not the AI: "Our artists explored 200 AI-generated concepts, painted the final 40 by hand" beats "We use AI in art." Never claim more human involvement than happened — the internet fact-checks.
4. Choose the venue and timing (input: statement → output: release plan)
   Disclose in your own channel first (dev blog, patch notes, store page section) — never let a data-mine, a leak, or a blunt executive thread be the reveal. Time it with something players like (a content drop, a free weekend), not buried in a Friday-night post. Prepare the FAQ before you publish: the five hardest questions, answered in your voice.
5. Triage the sentiment (input: launch → output: sentiment log)
   In the first 48 hours, sort reactions into: legitimate craft concerns (engage — these people might be right), training-data/job anxieties (acknowledge the industry problem honestly, state your studio's actual practice), bad-faith pile-ons (don't feed), and press inquiries (one spokesperson, written answers, no improv). Log volume and tone per bucket; decide at 48 hours whether a follow-up post is needed or silence is.
6. Handle the blunt-executive scenario (input: crisis → output: containment plan)
   If someone already posted the provocative version ("developers who refuse AI shouldn't be in business"): don't delete-and-pretend unless the account itself is the problem — deletion reads as discipline either way. Separate the person from the policy within 24 hours ("those were personal views; here's our actual practice"), publish the real inventory from step 1, and put the human-in-the-loop line in writing: what judgment stays human, no exceptions.

Anteprima del modello

AI-Use Disclosure Worksheet

Fill in before anyone asks — because someone will.

1. AI-use inventory (honest version)

2. Audience map

• Players fear ___. They need to hear ___.
• Press fears ___. They need to hear ___.
• Partners/platforms fear ___. They need to hear ___.

3. Disclosure statement (your words)

What was used / where in the pipeline / what humans did / what changed for the player:

Lead with the human work.

4. Release plan

• Own channel (dev blog / patch notes / store page):
• Timing (paired with what players like):
• Pre-written FAQ (the 5 hardest questions):

5. Sentiment triage (first 48h)

Follow-up post needed, or silence? Decide at 48h:

Il pacchetto della skill (SKILL.md + modello) è scritto in inglese.

Acquista il pacchetto completo della skill — $4.99 Pro — tutte le 511 skill