select-shots
Skill4 starsChanged 9 days agoMITView on GitHub
Review every clip in a footage folder and pick the most cinematic moments — sample stills from each clip, score sharpness, shake, exposure and colour, find each clip's steadiest window, judge framing (keeping heads in when portrait footage fills a landscape frame) and make sure every clip gets used. Use it before cutting any montage, recap, highlight or travel video from a folder of clips, whenever the user says "use the best shots", "make it cinematic", "you missed some of my videos", or complains a shot is shaky, blurry, too dark or too close.
In this repository6 other files it gives its agents.
- analyze-referenceSkill
- check-renderSkill
- filmcraft-editingSkill
- match-gradeSkill
- photo-titleSkill
- recap-editSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 52
Use every clip at least once.
- Distinctive: line 53
Prefer wide/medium shots full-frame.
- Distinctive: line 54
Keep a table: clip → moment → slot → focus → look.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Selecting shots
People remember their own footage. Two complaints come up again and again: "my best clip isn't in
it" and "why that moment?" (shaky, blurry, head cut off, a selfie blown up to fill the screen). Review
all the footage, every time, before planning the cut. Scripts need Pillow and
FILMCRAFT_CLI="${user_config.filmcraft_cli}".
1. Sample every clip
python3 ${CLAUDE_SKILL_DIR}/scripts/sample_clips.py <clips folder> work/samples --every 0.5 --skip recapStills every 0.5 s (long clips sparser), plus clips.json with each clip's length, size, fps and
whether it's portrait. Re-run it when the user adds clips — then diff against what the plan uses.
2. Score
python3 ${CLAUDE_SKILL_DIR}/scripts/score_frames.py work/samples --window 1.0Per clip: steadiest+sharpest window for a shot of --window seconds, median sharpness, motion
(high = shake or fast pan), brightness, and flags: dark, washed, shaky, soft, grey. Treat these as
pointers, not verdicts — a dark cave with a shaft of light, or a fast bike ride, can be the best shot.
3. Look
python3 ${CLAUDE_SKILL_DIR}/scripts/review_sheets.py work/samples work/review --aspect 16:9One block per clip, every still labelled with its time; on clips of a different shape a magenta box shows the slice that fills the frame when centred. Read every sheet. For each clip note:
- best moment(s) and why (light, gesture, action peak — a smash, a jump, arms opening at sunset)
- for portrait clips in a landscape edit: the focus height that keeps the subject in the slice (people: usually 0.3–0.45 so heads stay in; render a frame to confirm)
- whether it can fill the frame at all: selfies, tight close-ups and low-res (720p) portrait clips look "very zoomed" full-frame — plan them as whole pictures in an overlay section instead
- low light that will need denoising, 24 fps clips that need frame blending
- clips filmed upside down / sideways (rotate them)
For a promising moment, look closer at 0.1 s steps with export_frames(item, times) — e.g. to land
exactly on a smash or a jump.
4. Choose
- Holds (2–3.5 s) get the most cinematic, steadiest shots; half-beat bursts (0.35–0.45 s) can take busier moments; one hero shot per section.
- Use every clip at least once; repeat only long clips, at clearly different moments.
- Prefer wide/medium shots full-frame; save close-ups for overlays.
- Keep a table: clip → moment → slot → focus → look. It becomes the plan.
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.

