Find References
Find real, human-made references before anything is made, for AI image, video and music and for video made with code. Turns a request into one query pack per job (light first, then framing, people, location, texture, type, motion, palette, music and a "never" board), searches each source in its own words, opens everything it keeps, proves who made it, measures what code can measure, checks light with an LLM judge, and hands a person three candidates per job. Use when someone says "find references", "reference board", "what should this look like", "lighting reference", "framing reference", "tex
Install
/plugin marketplace add vargHQ/skillsThe repo is a Claude Code plugin marketplace. Run /plugin afterwards to pick the plugin that includes this skill.
npx skills add vargHQ/skills --skill find-referencesgit clone --depth 1 https://github.com/vargHQ/skills.git /tmp/skills
mkdir -p ~/.claude/skills
cp -r /tmp/skills/find-references ~/.claude/skills/find-referencesPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Find references > Step 1 of 10 in [How to make a video with an AI agent in ten steps](https://varg.ai/how-to/make-a-video-in-ten-steps). The full guide, with every test and source: https://varg.ai/how-to/find-references-for-an-ai-video (as Markdown: add `.md`). Without a reference an agent falls back to its default look: "centered text, gradient background, everything fading in" (https://x.com/rexan_wong/status/2103707054108299437). Adding more words does not fix it. Three facts shape this skill: - **No library searches by ratio, shadow edge or Kelvin.** Search with words a cataloguer would use, then measure what you find. A spec and a query are different things. - **Machines read light and taste badly** (`references/evidence.md`). A judge answers literal questions about what is visible, and nothing else. - **The best libraries forbid bots.** Film-Grab, ShotDeck, Pinterest, Instagram, Rate Your Music and Bandcamp's pages are for people. The agent reads open archives and writes links for the person. So each part does what it can do. **Code owns facts in pixels and timing. Records own who made it and the rights. The LLM owns literal yes/no checks and the shortlist. A named person owns meaning and the final pick.** **Some things only a person can find.** In the cold run on an ad (`tested 2026-11-03`), no open source returned the inside of an old Portuguese tile shop, the thing that person loved most. Finding that out early and writing the link is a
Details
- Skill name
find-references- Repository
- vargHQ/skills
- Path
find-references/SKILL.md- License
- MIT
- Repo updated
- 18 November 2026
- Listed in
- heilcheng/awesome-agent-skills
FAQ
What does the Find References skill do?
Find real, human-made references before anything is made, for AI image, video and music and for video made with code. Turns a request into one query pack per job (light first, then framing, people, location, texture, type, motion, palette, music and a "never" board), searches each source in its own words, opens everything it keeps, proves who made it, measures what code can measure, checks light with an LLM judge, and hands a person three candidates per job. Use when someone says "find references", "reference board", "what should this look like", "lighting reference", "framing reference", "tex
How do I install the Find References skill?
Claude Code plugin: /plugin marketplace add vargHQ/skills. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Find References popular?
It lives in vargHQ/skills, which has 38 GitHub stars and 7 forks, and is included in 1 curated skill list.
Related skills
Ad Creative Testing
38vargHQ/skills
Turn one winning ad into a creative testing plan for TikTok and Meta. Decodes why an ad won (hook, argument, buyer, format), plans concepts that Meta's Andromeda and TikTok treat as genuinely different (buyer × angle × format × first frame, not caption swaps), sets how many creatives to launch, the budget per test and when to judge, writes file and ad names that make results readable, reads hook rate, hold rate, CTR and CPA to find which part failed, and says when to kill, iterate or scale. Use when someone asks how many ads to test, how to make ad variants, why their variants all perform the
AI Ad Compliance Check
38vargHQ/skills
Turns an AI video ad into a compliance report for TikTok Shop, TikTok ads, Meta, YouTube, the FTC, New York's synthetic-performer law and the EU AI Act before it runs: varg video_analyze pulls every shot, spoken and on-screen claim and person, each claim is matched to the seller's evidence, the visuals are checked against TikTok Shop's AI product rules (no fake effects, before/after, changed quantity or size), people and voices against the FTC fake-testimonial rule, and the result is a pass / fix / block list with the rule, source and date, a rewrite for each fix, and the AI labels to set on e
AI Parody Video
38vargHQ/skills
Plan and make a short AI parody video: a known movie scene, TV format or news shot with one wrong element, 10 to 45 seconds, built to be shared. Gives the joke structure, six formats measured on clips with 60 thousand to 9 million views (scene recast with new dialogue, original-vs-remake, one-prompt sitcom sketch, fake live news shot, character monologue, silent slapstick), which production route each one takes, prompt templates for the sketch and the news shot, subtitle and caption rules, and a plain list of what is allowed with real people, known characters, studio footage and cloned voices,
Footage Effects
38vargHQ/skills
Eight effects for footage you already have, each as tested ffmpeg filter chains plus HyperFrames and Remotion equivalents: glow and bloom (soft, film halation, sci-fi, sunset, anamorphic streaks, ominous), digital glitch (subtle to corrupt), CRT screen (consumer TV, arcade grille, green and amber terminal), chromatic aberration (lens, radial, linear split), halftone (ink on paper, CMY screen, LED wall, mosaic), acid photocopy, radial zoom blur and wave or heat distortion. Presets give thresholds, blur scales, offsets in px at 1080p, screen angles, ink colors, scanline periods and where in the
Ecommerce Video Edit
38vargHQ/skills
Turn raw product footage into a TikTok ad with FFmpeg. Recuts creator, UGC or product footage into a 9:16 ecommerce ad for TikTok, TikTok Shop, Reels and Meta: the best moment moved to second 0, a text hook inside the safe zone, dead air cut, captions, a CTA end card, loudness at -14 LUFS, and one MP4 per hook for testing (plus a 4:5 feed cut and a cover frame). Every command was run on test media. Free and local; on the varg ffmpeg tool (credits) when the agent has no shell. Use when someone has footage of a product (a creator video, an unboxing, a demo, a winning ad) and wants it edited into
Gpt Image 2 6
38vargHQ/skills
Guide to OpenAI's GPT Image 2.6 on varg (gpt_image_2_6) and the entry point to the GPT Image 2.6 skill set: Flare vs Sunburst, quality vs reasoning_effort, sizes and aspect presets, what fal rejects (xhigh and max effort, SVG, over 16 GIF frames), measured prices, failure modes, the exact varg MCP and REST calls, and prompt recipes for the text-heavy layouts people shared on X: infographic, product launch poster, photo to editorial travel print, photo to Baroque gallery poster, landmark postcard, character turnaround sheet, storyboard sheet and low-res to clean upscale, each credited to its au
Browse the most popular Claude skills, all images, video & audio skills, or write your own with the skills generator.