agentsclimarketplace

Script to video production agent

Skill jackterror/script-to-video-production-agent

A modular AI-ready script-to-video production agent that turns tutorial scripts into narration-locked scenes, reviewable visual direction, prompt packs, and local asset queues.

Install
npx -y skills add jackterror/script-to-video-production-agent

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Turns tutorial or explainer scripts into narration-locked Scene blocks, human-reviewed visual direction, vendor-ready video-builder prompts, still and video prompt packs, image-only exports, and a resumable local asset queue. Use for DOCX, Markdown, or plain-text script imports, review-sheet generation, prompt-pack creation, queue seeding, retry tracking, release packaging, or clean-room script-to-video workflows that require human approval before visual changes or paid generation.

SKILL.md

4.3 KB, 841 tokens by cl100k_base, as published. Nobody here has run it

Script-to-Video Production Agent

Turn a source script into a controlled script-to-video workflow. Keep narration exact. Keep visual changes reviewable. Keep generation state resumable and local.

Operating Rules

  1. Preserve narration exactly by default. Treat narration edits as exceptional and deliberate.
  2. Use the Python runtime for parsing, hashing, rendering, queue state, and audit checks.
  3. Use the host agent for script judgment, visual judgment, prompt writing, and review reasoning.
  4. Require human approval before visual changes, paid generation, asset acceptance, or final delivery.
  5. Treat the SQLite queue as the source of truth for asset attempts and approval state.
  6. Do not claim a live provider integration unless it has been explicitly tested in the current environment.
  7. Run the release audit before calling a package public-ready.

Workflow

1. Import and normalize

Run:

python3 scripts/stv_agent.py import-script \
  --input /path/to/source-script.docx \
  --config assets/config/examples/physical-process.json \
  --project-dir runtime/project-name

The runtime reads DOCX, Markdown, or plain text and normalizes the script into ordered Scene blocks with Narration and Visuals.

Read references/scene-schema.md before changing the schema contract.

2. Audit visual direction

Run:

python3 scripts/stv_agent.py review-export \
  --project-dir runtime/project-name \
  --project-id fixture-physical

Review the generated Step 2 document. The host agent should reason about visual issues, but the exported decisions must use the explicit A, R, or F format.

3. Apply review decisions

Run:

python3 scripts/stv_agent.py review-apply \
  --project-dir runtime/project-name \
  --project-id fixture-physical \
  --decisions runtime/project-name/outputs/review-decisions.txt

This writes before/after traceability and a cleaned v2 script.

4. Export prompt documents

Choose the required output:

python3 scripts/stv_agent.py export-builder-prompt --project-dir runtime/project-name --project-id fixture-physical
python3 scripts/stv_agent.py export-invideo-prompt --project-dir runtime/project-name --project-id fixture-physical
python3 scripts/stv_agent.py export-visual-pack --project-dir runtime/project-name --project-id fixture-physical
python3 scripts/stv_agent.py export-image-only --project-dir runtime/project-name --project-id fixture-physical

Use references/prompt-profiles.md before editing any exported prompt shape.

5. Seed and operate the asset queue

Run:

python3 scripts/stv_agent.py queue-seed \
  --project-dir runtime/project-name \
  --project-id fixture-physical \
  --asset-kind image
python3 scripts/stv_agent.py queue-next \
  --project-dir runtime/project-name \
  --project-id fixture-physical

Record attempts, approvals, rejections, and retries through the queue commands. Read references/providers.md before representing a generation path.

6. Audit the release

Run:

python3 scripts/release_audit.py
python3 scripts/package_release.py

Read references/clean-room-release.md before preparing a public archive.

Resources

  • scripts/stv_agent.py: CLI entrypoint.
  • assets/config/project-profile.schema.json: profile contract.
  • assets/config/examples/: example configurations.
  • assets/fixtures/: synthetic source inputs and expected outputs.
  • references/scene-schema.md: canonical schema and narration-lock rules.
  • references/prompt-profiles.md: prompt-export contract.
  • references/providers.md: provider and approval contract.
  • references/higgsfield-mcp.md: optional Higgsfield integration guide.
  • references/clean-room-release.md: public-release requirements.

What ships with it: 53 files

213.4 KB alongside SKILL.md, 20 of them executable

agents/

assets/

docs/

evals/

references/

scripts/

src/

13 more files not listed here. See all 53 in the repository.

Gives 0 of the 12 instructions most context ai engineering skills give in 841 tokens

Counted across 1,193 of the 1,976 authors here whose files we hold, read 2026-08-07

  • Dispatch a fresh implementer subagent per taskin 48 of 1193, across 19 files
  • Dispatch a final code reviewer after all tasksin 33 of 1193, across 8 files
  • Provide full task text to the subagentin 30 of 1193, across 9 files
  • Review spec compliance before code qualityin 27 of 1193, across 10 files
  • Make the hook script executablein 26 of 1193, across 8 files
  • Re-snapshot after navigation or DOM changesin 25 of 1193, across 19 files
  • Read files before editing themin 22 of 1193, across 11 files
  • Answer subagent questions before proceedingin 22 of 1193, across 7 files
  • Mark task complete in TodoWrite after approvalin 22 of 1193, across 6 files
  • Merge hook into existing settingsin 21 of 1193, across 3 files
  • Ask if installation is global or projectin 20 of 1193, across 2 files
  • Copy the hook script to target locationin 20 of 1193, across 2 files

Said here and by no other author read

  • Preserve source narration exactly by default
  • Use Python runtime for parsing, hashing, and queue state
  • Use host agent for visual and review reasoning
  • Require human approval before paid generation
  • Treat SQLite queue as source of truth for assets
  • Run release audit before calling package public-ready

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 327,069. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.