--- name: prompt-library-campaign-framings description: "The adopted 'engineering-campaign' framings (owner approved 2026-07-04, from a ChatGPT prompt-suggestions review) — what was ADOPTED, what was FOLDED into existing audits, what was SCRAPPED and why. Plus the reusable-Library idea and the reason it must be gitignored." metadata: node_type: project type: project originSessionId: prompt-library-campaign-framings --- **Owner 2026-07-04: *"do all your recs, make sure the prompt draft is in a folder for easy access."*** After the owner ran the site and pre-website code through ChatGPT asking for prompt ideas, Dispatch reviewed the output. **The verdict, and the reason it matters:** **most of ChatGPT's list already mapped onto the existing parked audit chain** (diegetic, code-quality, performance, a11y, content, mobile, test-strength, UI-consistency, cloud → synthesis → master plan → parallelization). So: **ADOPT the genuinely-new pieces, FOLD the rest into existing audits, and do NOT build a parallel system or generate hundreds of prompts.** Analysis-paralysis risk; keep the build-to-plan ratio healthy. *Trimmed 2026-07-20: this file used to carry a full folder-structure and relocation spec for the Library. That spec no longer matches the repo — **look at the filesystem, not at this file**, for where the Library actually lives. The one durable part is kept below.* ## ⭐ WHY THE LIBRARY MUST BE GITIGNORED (the load-bearing reason — don't lose it) Two reasons, and the second is the one that bites: 1. **Privacy** — it must never land on the public GitHub repo. 2. **Survival** — the build sessions run `git clean -fd`, which respects `.gitignore`. **An untracked-but-not-ignored top-level folder would be WIPED by the session cleanup.** So gitignoring it is **mandatory, not optional.** The Library is a **LIVING doc** — grow it as you go, pull from it when launching a campaign. It outgrew "prompts" into general reference (framings, briefs, reusable content). ⚠ **Known hazard: it has previously existed in two places at once.** Before adding to it, confirm which copy is canonical. ## GENUINELY-NEW THINGS ADOPTED (owner: "do all your recs") 1. **The library artifact itself** — codify and reuse prompts instead of writing them ad hoc. 2. **Red-team / devil's-advocate workflow** — for big DESIGN/ARCHITECTURE decisions, an adversarial pass where the planning model argues AGAINST the plan before build. Extends plan→audit with an adversarial stage. **Use for high-stakes calls, not routine work.** 3. **Cross-System Interaction audit** ✅ (owner adopted first-class) — *"find systems that never talk but should"* (timeline / GPS / inventory / quests / companions / faction / vendors / weather / time / audio / AI / hardcore). Generative for EMERGENT gameplay; folds into the gameplay ideation and capability survey. ⚠ **Filter every proposed link through player-control — deterministic and opt-in wherever possible.** **The owner sees this as one of the highest-value ideas.** 4. **Assumption-Challenge campaign** ✅ (owner adopted first-class) — a PHILOSOPHY audit, not bugs or features: identify assumptions accepted without re-evaluation (naming, architectural boundaries, UX patterns, feature ownership, implementation order, abstractions, roadmap sequencing) that persist from **historical momentum and institutional inertia**; recommend changes only where there's clear long-term benefit. **Distinct from "surprise me"** (which targets improvements) — *this targets inherited decisions nobody re-examined.* Run roughly once per round. 5. **One open-ended "no predefined objective / surprise me / embarrass me / forget the roadmap" campaign PER ROUND** — a fresh-eyes pass to catch blind spots the structured audits miss. Time-boxed, analysis-only, feeds synthesis. **A periodic complement, NOT a replacement for the structured audits.** 6. **Future-proofing lens** (predict-future-problems / museum / modding / community / AI-first) → folds into code-quality + synthesis. ## FOLD-INS — map new framings onto EXISTING audits, do NOT duplicate - Repository archaeology / delete-code / magic-numbers / naming / readability → [[parked-code-quality-audit]]. - Terminal & AI personality / CRT / lore / environmental storytelling / boot-sequence expansion / terminal history / corporate bureaucracy → [[parked-diegetic-audit]] (which derives the house standard) + content/gameplay ideation. - Hidden-opportunities / cross-system → [[parked-capability-survey]] + [[parked-ai-feature-eval]]. - **"Think like Carmack / Valve / Nintendo / Apple / id" personas → framing LENSES only.** The goals they'd produce are already audit dimensions; low incremental value, **not new campaigns.** ## ❌ SCRAPPED (owner): "What would Bethesda do?" — strip everything that feels like a modern website **Directly conflicts with the design north star: modernize WITH immersion — diegetic AND genuinely good and usable.** The overhaul MODERNIZES with immersion as the focus; **it does NOT remove modern features.** The good kernel (audit interactions for immersion) already lives in the diegetic audit, framed correctly. ## 📋 ADRs (Architecture Decision Records) — ADOPT LATER (owner floated 2026-07-04: *"not now, later"*) Tiny half-page decision files (Title / Status / Problem / Decision / Consequences). **Why they aren't urgent:** memory ALREADY handles the institutional-memory FUNCTION — decisions plus rationale, auto-loaded into the orchestrator's context. **What ADRs genuinely ADD beyond memory:** (1) **discoverability for dispatched CODE sessions** — orchestrator memory does NOT auto-load into them, so a decisions folder plus a pointer would let any session look up *"why is it this way?"*; (2) a clean, canonical, human-auditable format; (3) **formal superseding** (Status: Superseded by …). ⚠ **CAVEAT: do NOT duplicate every decision into ADRs — drift risk.** Reserve ADRs for the roughly ten-to-fifteen **LOAD-BEARING "never re-litigate" decisions** (modernize-with-immersion, player-control, keep-the-RobCo-brand, per-game device forms, the dev-branch model, free-infra-only, AI-never-sole-authority…). The working design memory stays the live store for everything else. **When it happens (later, not mid-overhaul): back-fill the load-bearing decisions from memory and add a pointer from the constitution.** Pairs naturally with the Assumption-Challenge campaign — ADRs give it the list to challenge. ## 📊 MEASURE USER-FACING OUTCOMES, not just code quality (the one genuinely-new bit of that batch) Code quality is measured heavily; **UX outcomes are not measured at all.** ⚠ **HARD CONSTRAINT: this project deliberately has NO analytics or tracking** (personal project, privacy and ads-free ethos) — so **literal player-usage telemetry is OFF the table.** **The ADOPTABLE version:** bake outcome CRITERIA into DESIGN REVIEWS and the UX/archaeology audits — tap-count per common task, discoverability, learnability (the new-user path), and dead/unused-feature detection. **Fold these in as explicit evaluation criteria; no new tracking infra.** ## ALREADY PRACTICED — no new formalization needed **Proportional process / a lightweight path for small changes** is already how it works: small fixes go straight to a build-and-verify session; only non-trivial or design work gets the full plan→build→audit; only architectural work gets the whole audit chain — and the protocol already says the stages aren't rigid, *"Dispatch judges per situation."* **Over-formalizing tiers would itself be process creep.** Related: [[robco-audit-program]], [[parked-synthesis-masterplan]], [[parked-diegetic-audit]], [[parked-code-quality-audit]], [[parked-capability-survey]], [[design-overhaul-direction]].