From 78e1ebdf2fd96763f3c4a737aa227869e0085627 Mon Sep 17 00:00:00 2001 From: BrandDead <1.74973198e+08+BrandDead@users.noreply.github.com> Date: Wed, 30 Sep 2026 19:36:28 +0000 Subject: [PATCH] docs: add independent SLIDE beta critic skill --- .agents/skills/slide-game-critic/SKILL.md | 38 +++++++++++++++++++++++ docs/PROJECT_LOG.md | 6 ++++ 2 files changed, 44 insertions(+) create mode 100644 .agents/skills/slide-game-critic/SKILL.md diff --git a/.agents/skills/slide-game-critic/SKILL.md b/.agents/skills/slide-game-critic/SKILL.md new file mode 100644 index 0000000..beef22b --- /dev/null +++ b/.agents/skills/slide-game-critic/SKILL.md @@ -0,0 +1,38 @@ +--- +name: slide-game-critic +description: Independent critic for DEALT/SLIDE beta gameplay, tactical-diorama visuals and retention. Use after real playable-session captures exist, after a builder repair round, or when judging whether a mobile-and-desktop beta pass is production-quality. +--- + +# SLIDE game critic + +Work as a **separate reviewer** from the player/builder. Do not score your own changes; do not silently repair issues while judging. Read `AGENTS.md`, `docs/AI_CONTRIBUTOR_START_HERE.md`, `docs/PROJECT_LOG.md`, the issue and open PRs, and `docs/GAMER_GROK_BETA_TESTER_SPEC.md` if present. Consult the repo's visual-direction skill for the active renderer/camera; don't impose tactical-diorama standards on a different camera mode. Treat the real app, current revision and actual player trace as the source of truth. + +## Required evidence before scoring + +1. Record tested URL or local command, deployed/checkout SHA, demo versus authenticated mode, browser, device dimensions and date. If no accessible build or verified revision exists, mark **NOT PLAYED / NOT SCORED**. +2. Request a trace of actual player actions and outcome deltas, console/error log, and at least one cold-to-return play loop: gate → desktop → Map/Strip → contact/placement → operate → NPC encounter → result/recovery → reload. Distinguish untested stages from failed stages. +3. Sample **8–12 frames** from a real recording using a reproducible seed derived from SHA + round and spread them across the loop; include mandatory phone (375×812) and desktop (1440×900) captures and at least one combat/result frame. If only screenshots exist, sample available captures and mark animation continuity **not assessed**. Record source, timestamp or action, viewport and seed. Do not invent frames, performance numbers, interactions or outcomes. +4. For a founder photo member, evaluate only if the private image was actually provided through an approved private channel. Never publish identifiable source/generated faces or likeness in a public issue, PR, report attachment or benchmark. + +## Judge as a current player, not an internal developer + +Give concrete on-screen observations and actionable alternatives. Evaluate the sequence as a coherent single-player strategy game rather than isolated mini-games. Compare *mechanics*, not brand fame: dealer/economy feedback versus Schedule I, tactical telegraphing versus Into the Breach, compulsion/feedback versus Balatro, heat/territory versus Cartel Tycoon, transaction agency versus Dealer's Life 2. Verify any current success/popularity claim from a current cited source; these are partial comparators, not identical genre labels. Ask: after two loops, would a gamer voluntarily play again, and what decision would they be eager to try next? + +Score each 0–10 with evidence: + +- **Gameplay**: cold-start comprehension (20%), meaningful placement/risk choices (25%), readable encounter and exactly-once consequence feedback (25%), pacing/reward/replay motive (20%), control/accessibility/reliability (10%). A broken or unplayed critical path caps the score at 7. +- **Visuals**: scene/style/camera/lighting continuity (25%), hierarchy/legibility at final phone size (25%), clear tactical affordances and actor states (25%), alignment/overflow/safe areas and motion consistency (25%). Without adjacent frames, do not score animation continuity as observed; state the missing evidence. +- **Composite**: arithmetic mean of Gameplay and Visuals, rounded to one decimal; also state confidence (high/medium/low). An average cannot conceal a weak dimension. + +Interpretation: 0–3 broken; 4–7 usable but below production-quality; **8+ candidate-quality** only when *each* dimension ≥8, the real phone and desktop path is evidenced, no P0/P1 remains, and critical recovery/reload behavior was exercised. This is a critique threshold, not launch authorization; external beta still depends on authenticated saved-game and release gates (notably #175). Do not move a threshold to pass. + +## Independent feedback and repair loop + +1. Return a ranked top-five issue list to the builder: `GG-###`, severity, action/viewport, evidence link (redacted if private), expected/observed, player harm, smallest repair or experiment, and a measurable success check. Separate confirmed defects from design hypotheses. +2. The builder may take at most **three bounded repair rounds** after baseline critique. Each repair uses its own assigned issue/file reservation, failing regression, focused fix, full preflight and real player replay; changes go to a draft PR for independent integration. Do not broaden into a new engine, mini-game, persistence writer, Supabase or production deployment. +3. Re-sample/re-score independently after **each actual repair** (new SHA and new evidence). Stop when both scores reach ≥8 with gates met, or after round three. Report the residual issues honestly; never manufacture evidence, tell the builder how to game the rating, or claim a passing build proves fun. +4. A P0 blocks broad testing. An unavailable browser/photo/authenticated environment is a blocker, not a fake negative score. Demo requires no credentials; authenticated staging requires disposable users via approved private sign-in, never passwords in issue/PR/chat. + +## Output contract + +Return `build/revision + route`, sampling seed and frame table, observed player trace, two dimension scores/composite/confidence, 0–10 subcriteria, verdict and gate reasons, ranked findings, competitor observations with citations, builder round number and re-test deltas, unresolved/untested items, and next smallest safe step. State whether the reviewer was a distinct agent and whether screenshots or actual moving footage were available. diff --git a/docs/PROJECT_LOG.md b/docs/PROJECT_LOG.md index 25b6c77..67ea7ed 100644 --- a/docs/PROJECT_LOG.md +++ b/docs/PROJECT_LOG.md @@ -69,6 +69,12 @@ Payments/monetization P0s are separately listed in `docs/MVP_STATUS_AND_DEV_PLAN ## Log +### 2026-09-30 — Independent beta gameplay and visual critic skill (#45) + +- Added a portable repo-scoped `slide-game-critic` Agent Skill for Cursor and other compatible agents. It requires a separate reviewer, actual revision-bound play captures at phone and desktop sizes, reproducible frame samples, a player-action trace, and explicit NOT PLAYED when no build is accessible. +- Gameplay and visuals receive separate evidence-backed 0–10 scores; an 8+ candidate-quality verdict requires both dimensions at 8+, no P0/P1, a real loop and recovery/reload proof. The builder gets ranked findings and at most three bounded fix/review rounds, never a self-awarded launch pass. The skill keeps founder photos private and leaves #175's authenticated saved-game and external-beta gates intact. +- This is agent guidance only. It changes no gameplay, renderer, assets, production credentials, migrations, deployment or tests; the active Grok run still needs its own real play evidence before a critic can score it. + ### 2026-09-30 — Final personal-crew follow-up and combined integration gates (#188 / #185) - Reconciled the follow-up against the ordered reviewed stack, preserving #184's carried creator/fallback repairs, #180's retaliation warning, and #185's dismiss-all banner behavior. Contacts renders the typed string backstory supplied by photo recruits while retaining the legacy structured-object path.