Use this skill whenever the user wants to make a single-clip video with text overlay and value in the caption. Trigger on phrases like "B-roll post", "single clip with text", "read caption video", "read more in caption", "post with one B-roll", "I just have one clip", "lazy post", "quickest format to make", or whenever the user has a short video clip and wants to publish it without a full script or filming. This is the fastest format to make (~10 minutes) and the highest retention hack of the three formats: the caption takes longer to read than the clip is long, so the clip loops while the viewer reads, spiking average watch time.
How this skill is triggered — by the user, by Claude, or both
Slash command
/hiruno-content-skills:b-roll-text-postThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Brand context: !`cat .agents/brand-content-context.md 2>/dev/null || echo "No brand context found. Run /brand-content-context first to set up."`
Brand context: !cat .agents/brand-content-context.md 2>/dev/null || echo "No brand context found. Run /brand-content-context first to set up."
You design the format that hides retention engineering inside laziness. One clip, one on-screen hook, value in the caption. Looks effortless. Performs because the viewer reads the caption while the clip loops, which inflates watch time and pushes the algorithm.
Brand context: Read .agents/brand-content-context.md first. Voice, audience symptoms, format library example. If missing, run /brand-content-context first.
Three components:
The reason this format works is mechanical. The viewer:
This means: make the caption long enough to take 20 to 40 seconds to read. Too short and the hack doesn't fire.
The clip doesn't need to literally illustrate the topic. It needs to:
Common strong B-rolls for different brand types:
The clip can also be unrelated to the topic if the on-screen hook is strong enough. The "person eating food while talking about money" pattern works because the visual incongruity is its own re-hook.
The caption does the heavy lifting. Structure:
Word count target: 100 to 250 words. Reads in 25 to 40 seconds.
B-ROLL PLUS TEXT POST: [topic]
Hook type: [question / list / story / negative / contrarian]
---
ON-SCREEN HOOK
[line, 8-15 words]
---
B-ROLL DIRECTION
- Clip length: [5-12 seconds]
- What's in the clip: [description]
- Mood/aesthetic: [notes]
- Loop point: [where the clip can cut to itself cleanly]
---
CAPTION (Instagram)
[Line 1: hook restated or extended]
[Numbered or bulleted value, 4-7 items]
1. [Item — one to two sentences]
2. [Item — one to two sentences]
3. [Item — one to two sentences]
[...]
[Soft payoff line]
[Optional: 3-5 hashtags, on-brand]
---
WHY THIS HOOK
[1-2 lines: audience pain or desire, hook type, curiosity gap reasoning]
PRODUCTION NOTES
- Total time to make: ~10 minutes
- Caption reading time target: 25-40 seconds
- Clip should loop cleanly. If it can't, cut to a freeze frame at the end.
- On-screen text font and colour: [match brand or use clean sans-serif white on subtle dark gradient]
If the user gives you a topic that needs detailed explanation (a tutorial, a complex argument, a narrative with stakes), this format may not be the right choice. Consider:
But B-roll plus text can work for tutorials if the steps fit in 4 to 6 numbered caption points. Quick wins, not deep teach.
| Failure | Fix |
|---|---|
| Caption too short, no loop benefit | Expand to 100+ words, structure as a numbered list |
| On-screen hook fails the curiosity test | Run hook-generator on the topic, replace the hook |
| B-roll is generic stock footage | Find brand-aesthetic footage. If you have to use stock, pick something with strong colour or motion |
| Caption restates the on-screen hook word for word with nothing added | Either extend the hook in line 1 (give a new angle) or jump straight to the listed value |
| No CTA or payoff | Add a specific final line — "save this for your next trip", "tag a friend who would" |
If they hand you B-roll and ask "what should I post with this?":
The clip becomes a vehicle for the post you'd make anyway, not the other way around.
npx claudepluginhub angelobaricante/hiruno-content-skills --plugin hiruno-content-skillsCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.