# Research Provenance `director-skills` was distilled from a private local research library covering AI-assisted story writing, cinematic language, image prompting, video prompting, continuity, and model adaptation. The private source guides are not redistributed in this repository, and their filenames are intentionally omitted. ## Source Categories Reviewed - Short-story structure and AI-assisted fiction workflows. - Short-film development, loglines, treatments, beat sheets, and screenplay support. - Fountain screenplay formatting and scene-writing workflows. - Cinematography, camera movement, lens language, lighting, color, texture, and realism. - Shot-listing, storyboarding, visual breakdowns, and production packaging. - Character, location, prop, costume, style, and reference-image continuity. - Cinematic image prompting for stills, reference sheets, keyframes, posters, props, and environments. - Cinematic video prompting for text-to-video, image-to-video, time-blocked shots, audio, motion, and continuity. - Model-specific prompt adaptation for contemporary image and video systems. - Prompt iteration, output diagnostics, failure modes, and quality-control loops. - Safety, provenance, rights, likeness, and release-readiness considerations. ## Distillation Policy - Workflows were rewritten as original operational guidance. - Long passages and prompt libraries were not copied. - Unsafe bypass-oriented material was excluded. - Model capabilities are marked with confidence levels or `unknown` where the source material did not clearly support a claim. - Legal, copyright, provenance, likeness, and rights notes are practical risk guidance, not legal advice. ## Maintenance Notes When adding new research: 1. Extract repeatable procedures, not long summaries. 2. Update the relevant skill references, templates, examples, and tests. 3. Keep private filenames and local paths out of public documentation. 4. For model behavior, separate official/current evidence from community or anecdotal evidence. 5. Mark time-sensitive or uncertain capabilities as `unknown` rather than guessing.