Claude Skills by badgids

Build and use per-sequence project context shards so an agent can work on a feature film without loading the whole project state.

Audit canon, story, screenplay, production documents, and generation prompts for identity, wardrobe, prop, geography, knowledge, time, lighting, audio, and dialogue continuity conflicts.

Stress-test a story, book, screenplay, visual concept, audio plan, film plan, trailer, campaign, or production approach through a rigorous dependency-aware interview before execution.

Synthesize the current Story-Film conversation and project state into a durable creative and production specification without reopening decisions that have already been settled.

Run bounded creative review using Critique-Correct-Verify or Debate-Judge for consequential story, screenplay, directing, storyboard, or shot decisions, with explicit criteria, persistent verdicts, and escalation instead of endless loops.

Run a structured creative or production interview that exposes assumptions in dependency order for stories, books, images, audio, video, films, trailers, campaigns, and production workflows.

Probe finished film, trailer, social, and audio deliverables against explicit delivery specifications for file existence, streams, codecs, resolution, frame rate, duration, audio properties, checksums, and optional signal warnings.

Define a reusable visual design grammar for film titles, posters, key art, lower thirds, thumbnails, social assets, press materials, and programmatic motion graphics using approved references, exact text, safe zones, and accessibility constraints.

Keep exact approved dialogue audio, hashes, timing, visible-sync intent, generation conditioning, and review audio aligned.

Measure approved dialogue before expensive generation and block impossible clip timing without silently changing speech.

Create a reusable voice bible and exact dialogue cue list from screenplay lines, including character voice identity, delivery intent, language, timing, and continuity across scenes.

Create a director book that turns screenplay scenes into performance, blocking, visual, camera, rhythm, transition, and sound intentions before individual shots are chosen.

Run a decision-tree creative interview while maintaining durable Story-Film decisions, terminology, canon implications, evidence links, and open questions as the discussion progresses.

Perform non-destructive deterministic editorial assistance including silence mapping and jump cuts, optional local transcription, subtitle burn-in, exact clipping, subject-aware reframing, and delivery compression.

Create a post-production edit plan and cue sheet from screenplay, shot list, generated clip durations, dialogue, music, and SFX, including transitions, pickups, and continuity risks.

Export canonical film editorial state as editable Kdenlive and/or Shotcut projects using a shared multitrack MLT-based manifest, target-specific serializers, project-relative media, effects/transitions, subtitles, and structural validation.

Compile a standalone portable editorial package with timeline events, shot media, dialogue, subtitles, music, sound cues, stems, placeholders, missing media, and interchange-ready records.

Reconcile selected shots, sequence order, and main timeline placements before a feature-film master is accepted.

Build and validate a durable CLAIM ledger for documentary, historical, technical, educational, press-kit, festival, and campaign facts, preserving sources, uncertainty, disagreement, adoption state, and downstream use.

Use the installed FFmpeg and FFprobe capability surface for deterministic video, audio, subtitle, stream, codec, container, filter-graph, metadata, capture, analysis, QC, and delivery operations without hardcoding build-specific features.

Audit the final feature-film project for required control artifacts, approved sequences, coverage, continuity, editorial, generation, master, and release evidence.

Render and verify the complete finished movie from the executable picture timeline, synchronized audio master, subtitles, and delivery specification.

Plan feature-scale local generation by declared RAM, VRAM, job dependencies, and resident model groups before heavy ComfyUI execution.

Compatibility redirect for pre-v0.0.31 projects. Story-Film generation is workflow-first; use generation-workflow-setup to select complete ComfyUI workflows instead of asking the user to assemble model/resource stacks.

Compile story, screenplay, continuity, shots, references, dialogue, score, and sound needs into model-neutral JSONL generation briefs ready for model-specific adapters.

Discover complete ComfyUI workflows only from Story-Film's extension-local comfyui_workflows library, then let the user select each required task from an ordinary numbered list.

Build a safe step-by-step human procedure for Story-Film production actions the agent cannot perform itself, with progress, confirmation gates, resumability, and secret-safe handling.

Write MiniMax H3 prompts for T2VA, I2VA, FL2VA, L2VA, and Ref2VA using the required audiovisual fields, keyframe alignment, speaker IDs, dialogue tags, reference labels, and timing.

Adapt MiniMax H3 prompts for surreal single-scene clips where rough hand-drawn animation physically contacts, deforms around, or interacts with live-action hands, objects, surfaces, and camera motion.

Use ImageMagick 7 for deterministic still-image and image-sequence conversion, geometry, composition, masking, color, text, drawing, analysis, comparison, animation, metadata, batch work, and production artwork while respecting runtime delegates and security policy.

Create a current Generation 5 Kdenlive .kdenlive project from canonical editorial state using Kdenlive document version 1.1, main_bin, sequence and track tractors, MLT playlists, source producers, optional effects/transitions, and portable media references.

Adapt still-image briefs for Krea 2, using exploratory prompting for open visual search and locked prompting plus style references for production-consistent look development.

Deterministically snapshot, unload, verify, and restore models on a local llama.cpp llama-server router or Ollama server during resource-safe ComfyUI generation without asking the LLM to invent lifecycle scripts.

Track and verify continuity facts that must remain consistent across distant feature-film sequences.

Adapt shot briefs into LTX 2.5 prompts with shot scale, scene and light, physical action, character detail, camera movement, dialogue, ambience, music, and explicit cut continuity for multi-shot generations.

Plan and generate film key art, posters, thumbnails, title cards, social stills, banners, and platform-safe promotional images from approved visual identity and verified text.

Safely delete explicitly rejected registered media and repair missing disposable runtime copies from approved project media.

Review generated candidate takes with structured script-faithfulness, identity, background, spatial, action, motion, physics, artifact, dialogue-sync, and subtitle-sync checks before creative take selection.

Route deterministic media editing and manipulation through installed FFmpeg, FFprobe, MLT/melt, and ImageMagick runtimes with runtime capability discovery, portable manifests, safe argv execution, and reproducible evidence.

Adapt MiniMax H3 prompts for minimalist physical-product ads with strict product fidelity, clean negative space, product-specific motion, restrained integrated typography, and premium camera language.

Adapt Story-Film shot briefs into MiniMax H3 audio-video prompts using the mandatory H3 prompt-writing format plus an appropriate optional style skill for T2VA, I2VA, FL2VA, L2VA, or Ref2VA.

Adapt score or song cues into MiniMax Music 3 prompts using vivid English musical sentences plus explicit genre, mood, vocals, instruments, BPM, lyrics, and structure fields.

Export the executable timeline as portable MLT XML for generic MLT interchange while keeping the FFmpeg-rendered film master as the standalone delivery path.

Use the installed MLT framework and melt command for multitrack service graphs, producers, playlists, filters, transitions, links, consumers, XML serialization, playback, rendering, and editor-compatible timeline work with runtime service discovery.

Plan and render reusable film and campaign motion graphics such as title cards, lower thirds, watermarks, end cards, text overlays, framing treatments, fades, and transitions using stable GFX identities and verified text.

Adapt MiniMax H3 prompts for music videos with locked master audio, exact lyrics, beat-aware shot timing, spatial lyric typography, performance continuity, and multi-clip stitching when duration exceeds one H3 generation.

Adapt MiniMax H3 prompts for tactile paper-collage explainers using halftone cut-paper metaphors, layered compositions, readable narration beats, stop-motion paper movement, and tactile sound.

Adapt MiniMax H3 prompts for educational papercraft stop-motion using layered paper dioramas, visible physical construction, multi-plane depth, tactile paper mechanisms, stepped motion, and readable explanation.

Inspect, extract, render, merge, split, rotate, search, and repair PDF production documents through safe local tools, with optional MuPDF mutool support and permissive fallbacks when equivalent.

Convert approved screenplay lines or beats into playable performer staging with initial state, movement, actions, end state, props, timing, and capability-aware constraints.