agentleFS
badgidsbadgids/Story-Film-Skills/skills/story-film/SKILL.md

story-film

Skill4 starsChanged 3 months agoApache-2.0View on GitHub

Route story writing, books, screenwriting, directing, creative decision discovery, production specifications and work units, narrative state, creative review, storyboards, executable production, resource-safe AI media generation, generalized asset approval, audio mastering, timeline rendering, finished-film mastering, trailers, social campaigns, deterministic media editing, production documents, Kdenlive/Shotcut project export, and release packaging through a standalone file-based story-to-screen-and-release pipeline.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. About 0 are common, 9 are distinctive, and 9 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 9

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When the user explicitly invokes Story-Film, this router, its selected playbook, and 00_project/pipeline_progress.json are the Story-Film …

    line 13
  2. Very specific:

    Read CATALOG.md beside this file.

    line 23
  3. Very specific:

    Read that playbook in full from skills/story-film/playbooks/<playbook>.md in a Git package.

    line 25
  4. Very specific:

    If the project already exists, or it was initialized without --playbook, read ../pipeline-progress/SKILL.md and run …

    line 27
  5. Distinctive:

    Do not begin story, canon, screenplay, preproduction, or generation-brief work while a required ComfyUI workflow preflight is incomplete.

    line 29
  6. Distinctive:

    If ComfyUI is unavailable or a required workflow cannot be selected, mark the workflow-preflight target blocked.

    line 30
  7. Distinctive:

    If a matching progress ledger already exists, resume its current target instead of reconstructing progress from chat history or replacing …

    line 31
  8. Very specific:

    Update 00_project/state.json after each completed artifact.

    line 34
  9. Very specific:

    If an approved upstream artifact changes, run project-impact before rebuilding downstream work.

    line 35
  10. Distinctive:

    Run continuity, narrative-state, production-coverage when applicable, dramaturgy, prompt, standalone, and style checks at the gates …

    line 36
  11. Distinctive:

    Do not invent a new workflow while a listed playbook fits.

    line 40
  12. Distinctive:

    If the host provides a generic Todo tool, do not duplicate the full Story-Film playbook there.

    line 46
  13. Very specific:

    If the requested endpoint is a story bible, deep character/world lock, or durable cast/world development, route through …

    line 54
  14. Very specific:

    Before any attempt to locate ComfyUI, determine whether its server is running, inventory installed models, search model filenames, inspect …

    line 61
  15. Distinctive:

    Every PDF, DOCX, DOTX, XLSX, XLSM, XLTX, PPTX, PPTM, ODT, ODS, or ODP created by this suite must have a meaningful Markdown file with the …

    line 81
  16. Very specific:

    If the request is about FFmpeg, FFprobe, MLT/melt, ImageMagick, Kdenlive project export, Shotcut project export, or deterministic media …

    line 85
  17. Distinctive:

    If the requested final deliverable is an actual ComfyUI inspection, workflow run, generated asset, queue action, or ComfyUI repair, choose …

    line 89
  18. Distinctive:

    If the user names a specific artifact or model, route directly to the corresponding specialist inside the chosen playbook when the supplied …

    line 93

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Story Film

This is the general entry point for Story-Film Skills. The package is standalone.

When the user explicitly invokes Story-Film, this router, its selected playbook, and 00_project/pipeline_progress.json are the Story-Film execution authority. Do not substitute stage sequences, pipeline commands, or routing rules from unrelated installed skill packs unless the user explicitly asks to combine them.

Resolve every relative path in this skill from this SKILL.md, not from the user's project directory or the package root. CATALOG.md is the file beside this router (skills/story-film/CATALOG.md in a Git package). Every playbooks/<name>.md path is relative to this router, so its Git-package path is skills/story-film/playbooks/<name>.md; never try <package>/playbooks/<name>.md. The listed router, catalog, playbooks, scripts, references, and sibling skills are known package paths. Do not use Bash, find, ls, or directory scans to rediscover them.

Start

  1. Read ../../references/STANDALONE_CONTRACT.md.
  2. Read ../../references/CORE_CONTRACT.md.
  3. Read ../../references/DOCUMENT_COMPANIONS.md.
  4. Read ../../references/WORKFLOW_SELECTION.md.
  5. Read CATALOG.md beside this file.
  6. Match the request to exactly one playbook in playbooks/.
  7. Read that playbook in full from skills/story-film/playbooks/<playbook>.md in a Git package.
  8. For a new project with a multi-step playbook, initialize the project and authoritative progress ledger atomically with ../../scripts/init_story_project.py <project> [--title TITLE] [--format FORMAT] --playbook <playbook>. Do not create pipeline_progress.json or workflow_preflight.json manually.
  9. If the project already exists, or it was initialized without --playbook, read ../pipeline-progress/SKILL.md and run ../../scripts/pipeline_progress.py init <project> --playbook <playbook> before writing any story, canon, screenplay, preproduction, or generation artifact. A film/video project with an inactive progress ledger is not ready for creative production.
  10. Pipeline initialization for a mapped ComfyUI-backed playbook creates the matching workflow-preflight state and makes generation-workflow-setup the first current target. Complete that target before any story or canon artifact is created; use generation-workflow-setup for every missing category until workflow_preflight.py status reports complete.
  11. Do not begin story, canon, screenplay, preproduction, or generation-brief work while a required ComfyUI workflow preflight is incomplete. Later playbook stages reuse the durable selections and do not ask again unless the user explicitly requests a workflow change.
  12. If ComfyUI is unavailable or a required workflow cannot be selected, mark the workflow-preflight target blocked. Do not continue into creative work and do not promise to return to preflight later.
  13. If a matching progress ledger already exists, resume its current target instead of reconstructing progress from chat history or replacing the ledger.
  14. Execute the playbook in order. Before each specialist step, read the named sibling SKILL.md.
  15. After each actionable progress leaf, validate its artifact before checkpointing it complete. A blocking validation failure must remain on the same leaf and become blocked; it must not advance. Do not start a later specialist or write a later artifact until the current leaf is validated and checkpointed.
  16. Update 00_project/state.json after each completed artifact.
  17. If an approved upstream artifact changes, run project-impact before rebuilding downstream work.
  18. Run continuity, narrative-state, production-coverage when applicable, dramaturgy, prompt, standalone, and style checks at the gates required by the playbook.

Routing rule

Do not invent a new workflow while a listed playbook fits. If more than one fits, choose the one whose final deliverable most closely matches the user request.

Small-model rule

Work one artifact at a time. Save it. Validate it. Checkpoint the corresponding pipeline leaf. Then continue. For long work, process one chapter, scene, sequence, or stale dependency slice at a time. After restart or compaction, read 00_project/pipeline_progress.json and 00_project/HANDOFF.md before opening unrelated context.

If the host provides a generic Todo tool, do not duplicate the full Story-Film playbook there. Mirror at most three Story-Film items: current target, immediate next target, and requested endpoint. Refresh that mirror only after the authoritative Story-Film checkpoint advances.

Standalone rule

Never stop required creative or preproduction work because another skill pack is missing. Create the native portable artifact defined by this suite. ComfyUI is optional for planning and packaging. When the user asks to operate or generate through an available ComfyUI server, route through the native comfyui capability rather than assuming another agent extension is installed.

Creative planning routing

If the requested endpoint is a story bible, deep character/world lock, or durable cast/world development, route through playbooks/story-bible-development.md. Story-Film uses its existing distributed brief/canon/story/character/world/state artifacts instead of creating a project-specific skill as a second canon database.

If the user wants to pressure-test an idea, resolve creative ambiguity, turn existing discussion into a durable production specification, split a large production into executable work units, or chart a project whose route is not yet clear, route through playbooks/creative-planning-and-execution.md. Facts that can be discovered from files or tools are agent work; creative decisions remain the user's unless explicitly delegated.

ComfyUI discovery precedence

Before any attempt to locate ComfyUI, determine whether its server is running, inventory installed models, search model filenames, inspect nodes, or discover workflows, read ../comfyui/SKILL.md and ../comfyui-discover/SKILL.md.

In Pi, story_comfy owns ordinary interactive live ComfyUI discovery. Do not use Bash, find, ls, which, locate, guessed personal paths, home-directory scans, model-folder scans, direct comfy-cli discovery commands, or raw ComfyUI HTTP calls for those facts. A failed guessed path or empty guessed folder proves nothing. If the managed Story-Film control surface fails, report that failure and use only the documented deterministic fallback from the ComfyUI skills.

Generation workflow selection

Workflow choice is front-loaded. When the chosen playbook will require ComfyUI, complete workflow_preflight.py and generation-workflow-setup before creative production starts. Record every required task workflow in 00_project/workflow_preferences.json. Later stages consume those selections without reopening the interview.

The selected workflow owns its concrete models, VAEs, encoders, LoRAs, audio models, upscalers, sampler settings, and other graph configuration. Reopen the numbered list only when the user explicitly requests a workflow change. A later dependency failure is a blocker, not authority to reselect silently.

Production-integrity routing

For reference-driven generation, preserve REF-### authority scopes and run reference-authority. When adjacent approved shots need motion continuity, use temporal-continuity; previous-shot tails are visual-only and cannot import prior audio. When approved dialogue audio exists, use dialogue-audio-authority and dialogue-timing-preflight before expensive visible-speech generation. For ComfyUI reference graphs, use comfyui-binding-audit after staging or conversion. Physical cleanup of rejected media or repair of a disposable runtime copy routes through media-lifecycle, never directory sweeps or regeneration of an approved source.

Resource-safe local generation routing

If the local LLM and ComfyUI cannot safely coexist in RAM or VRAM, route generation through playbooks/resource-safe-comfyui.md. Use llm-model-lifecycle for native llama-server or Ollama unload/restore. All prompts, workflows, uploads, parameters, dependencies, output destinations, and validation decisions must be finalized before the local LLM is unloaded. While the LLM is unavailable, only deterministic runners and the Pi progress extension may advance or report generation state.

Rich-document companion rule

Every PDF, DOCX, DOTX, XLSX, XLSM, XLTX, PPTX, PPTM, ODT, ODS, or ODP created by this suite must have a meaningful Markdown file with the same basename beside it. A pointer-only Markdown file does not satisfy the rule.

Deterministic media editing routing

If the request is about FFmpeg, FFprobe, MLT/melt, ImageMagick, Kdenlive project export, Shotcut project export, or deterministic media manipulation, route through playbooks/media-editing-and-project-export.md. Discover the installed runtime before using optional codecs, filters, delegates, devices, MLT services, or hardware features.

ComfyUI execution routing

If the requested final deliverable is an actual ComfyUI inspection, workflow run, generated asset, queue action, or ComfyUI repair, choose the matching ComfyUI playbook. Live ComfyUI facts must be discovered at runtime.

User control

If the user names a specific artifact or model, route directly to the corresponding specialist inside the chosen playbook when the supplied upstream inputs are sufficient.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.