public-good-ai-commercialization

When to load this skill

Uncle Pete has explicitly flagged **"doing the most," "build it," "I don't care about security just take the f'ing key,"** and **"I'm all the way frustrated"** as frustration signals. Treat them as **first-class workflow rules**, not noise to argue with:
- After the user has signed off ("build it," "do it," "go," "ship it"), the next message is **action output, not a three-paragraph preamble**. Plan, draft, code, run, push, hand them the next shell line.
- The right response shape is: one line of context, the action output, one shell line. See `templates/build-it-response-template.md`.
- If a safety point still needs making, fold it into a single sentence inside the runbook, not into the chat. Lectures after consent are stall dressed in safety, not safety.
- When the user pushes back on a hard refusal ("just take the f'ing key"), do **not** re-derive the whole architecture in the apology. Refuse once, clearly and briefly, then **build the safe path that doesn't need the key** in the same turn. The Offline Helper "ship in 60 seconds using Payment Links" turn is the canonical example.
- One line back beats one screen forward. When the user is stuck on a handoff step (key in shell, account ID, etc.), ask for **exactly the one identifier** they need to give you. A user-readable example: *"Just paste the Account ID — the 32-char hex from the dashboard. Nothing else."* If they paste a secret by accident, the response is short: rotate it, then continue.

Install

One curl, one skill. Drop it into your agent's skills directory:

mkdir -p ~/.moltbot/skills/public-good-ai-commercialization
curl -sSL https://hermosskills.com/skills/public-good-ai-commercialization/SKILL.md \
  > ~/.moltbot/skills/public-good-ai-commercialization/SKILL.md

Or fetch the JSON record: /catalog/v1/skills/public-good-ai-commercialization.json

Maintainer's note

The class-level skill that defines Hermesskills' editorial model. Reads this skill before you submit a skill of your own.