prompt-lens-video-remix
Skill698 starsChanged 3 months agoMITView on GitHub
Analyze a user-provided reference video locally, detect shots, extract representative frames, reverse-engineer generation prompts, and optionally rewrite the concept or create a remake plan. Use when the user asks to break down, reverse-prompt, rewrite, or recreate a video; do not use for ordinary playback or unrelated writing.
In this repository10 other files it gives its agents.
- speckit-analyzeSkill
- speckit-checklistSkill
- speckit-clarifySkill
- speckit-constitutionSkill
- speckit-convergeSkill
- speckit-implementSkill
- speckit-planSkill
- speckit-specifySkill
- speckit-tasksSkill
- speckit-taskstoissuesSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Prefer the current agent's native visual and reasoning capabilities.
- Distinctive: line 13
Never install software, download a linked video, upload media, call a paid API, or start video generation without telling the user what …
- Distinctive: line 14
Never request an API key in chat or write one into project files.
- Distinctive: line 16
Keep generated frames and reports in a new user-visible output directory.
- Distinctive: line 20
Read references/workflow.md and select the local-file or link-input path.
- Very specific: line 21
Run python scripts/check_environment.py.
- Very specific: line 30
Read manifest.json, then inspect every extracted frame.
- Distinctive: line 32
Ask whether the user wants to keep the analysis, rewrite a new story, or create a same-structure remake.
≈ “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.
Files
Prompt Lens: Video Analysis & Prompt Reverse Engineering
Analyze a reference video shot by shot, automatically detect its real edits, and reverse-engineer a generation-ready prompt for every scene while keeping media local by default. After the factual breakdown, optionally rewrite the story or create an original remake using the same shot structure.
Operating rules
- Prefer the current agent's native visual and reasoning capabilities. Offer a user-selected API only when native capability is insufficient.
- Never install software, download a linked video, upload media, call a paid API, or start video generation without telling the user what will happen and obtaining approval for that action.
- Never request an API key in chat or write one into project files. Use the agent's secret store or a provider-standard environment variable configured by the user.
- Operate only on the media the user identified. Do not bypass DRM, authentication, private access, paywalls, or platform controls.
- Keep generated frames and reports in a new user-visible output directory. Do not overwrite or delete existing files.
Workflow
- Read references/workflow.md and select the local-file or link-input path.
- Run
python scripts/check_environment.py. Prefer the project's tuned PySceneDetect AdaptiveDetector. If it is unavailable, explain that FFmpeg0.32is the fallback and ask whether to use the fallback or install PySceneDetect. Do not run an installer yourself without approval. - Resolve the input to a local media file. For link handling, follow the boundaries in
references/workflow.md. Run:
python scripts/prepare_video.py INPUT --output OUTPUT_DIRECTORYThis produces
manifest.json,storyboard.md, and three representative frames per analysis segment. By default it reuses the project settings: PySceneDetect AdaptiveDetector with adaptive threshold3.0, minimum content value15.0, window width2, and minimum scene length0.6seconds; FFmpeg scene threshold0.32is the fallback. The manifest records the detector and distinguishes detected shot groups from duration-only subdivisions.- Read
manifest.json, then inspect every extracted frame. Analyze each scene with references/shot-analysis-schema.md. Treat motion, sound, dialogue, identity, and off-frame events as uncertain unless the available evidence supports them. - Present the overall structure, shot table, and generation prompts before proposing further action.
- Ask whether the user wants to keep the analysis, rewrite a new story, or create a same-structure remake. For either creative branch, read references/prompt-and-remix-guide.md.
- Generate actual media only when the user explicitly asks and confirms the selected tool/provider and any cost.
Completion standard
Deliver a concise overall reading plus a complete scene-by-scene result with timestamps, visual evidence, camera and motion notes, confidence labels, and a standalone generation prompt for each scene. Preserve the original analysis alongside any rewritten version so the user can compare them.
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
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.

