Use
Review the timing table first and validate anticipation and safe space in a simple blockout. Test final hit windows in-engine under real latency.
Design readable boss anticipation, active and recovery phases, safe space, and counterplay using redundant cues.
Design a {{attack_pattern}} for the {{boss_archetype}} described in {{boss_brief}}, viewed from {{camera_view}}, targeting {{difficulty_target}} readability.
1. Build a causal sequence: ready pose → primary anticipation → commitment point → active hit window → residual danger → recovery → counter window. Keep fake-outs as a separate proposal.
2. Redundantly signal with posture, weapon path, ground shape, luminance/scale change, particle direction, audio notes, and haptics rather than color alone. Keep danger direction and safe space visible from gameplay view.
3. Include one player only as a scale reference without forcing a hit. Lock camera, arena, boss proportions, and lighting; show one pattern per video.
4. Provide the final video prompt plus a timing table for phase, observable cue, gameplay hit state, player choice, cancelability, counter window, and accessible alternative cue.
5. Test readability without sound, at low saturation, on reduced-effects settings, and at close camera distance. Give acceptance criteria and playtest questions for fairness, silhouette, safe space, and recovery.
Shared execution rules
- First restate the inputs and selectors as a work contract; label unresolved details as explicit assumptions.
- Prefer current official documentation for engine features and separate fact, inference, and recommendation. If browsing is unavailable, do not claim recency.
- Respond in the order analysis → architecture → deliverable → verification table, with testable acceptance criteria for every section.
- Do not copy protected characters, games, or a studio's distinctive expression. Generated video and images are production references, not engine-ready assets.
- Rights, rigging, collision, root motion, networking, performance, and accessibility require human validation in the actual engine and target hardware. Never invent measurements or APIs.Avoid color-only cues, visuals that disagree with hit logic, hidden safe space, overexposure, clutter, gratuitous camera shake, mixed patterns, and forced hits.Review the timing table first and validate anticipation and safe space in a simple blockout. Test final hit windows in-engine under real latency.