Hidden Truth Reveal Video Gen

Create vertical reveal shorts: polished product vs harsh origin, one-line verdict, timed English captions (WeryAI). Use for supply-chain shock, contrast edit...

Security Scan

The skill's requirements and included code match its stated purpose (Node CLI that calls WeryAI video APIs using a single WERYAI_API_KEY); nothing requests unrelated credentials or system access.

Purpose & Capability

Name/description: video short generation (reveal/contrast) — declared requirements (node, WERYAI_API_KEY) and the included CLI (scripts/video_gen.js) all align with that purpose. Required binaries and env var are proportional and expected for a client that talks to WeryAI.

Instruction Scope

SKILL.md instructs the agent to expand prompts, confirm with the user, and run the included Node CLI which sends prompts/images to WeryAI. The instructions do not request unrelated files, credentials, or system state beyond the declared WERYAI_API_KEY and Node runtime. They explicitly document API hosts used.

Install Mechanism

No install spec (instruction-only) but the skill bundles an executable script (video_gen.js). That is low-risk compared with arbitrary downloads, but installing/running the included script will execute code locally and make outbound HTTPS requests to api.weryai.com and api-growth-agent.weryai.com — verify you trust those endpoints and review the shipped script before running.

Credentials

Only a single credential is required (WERYAI_API_KEY) and it is used as the bearer token for the documented API hosts. No unrelated secrets or config paths are requested.

Persistence & Privilege

The skill is not always-enabled and does not request system-wide persistence or modify other skills. It uses ordinary autonomous invocation (platform default), which is expected for a runnable skill.

Assessment

This skill appears coherent: it runs a bundled Node script that submits prompts and (https) image URLs to WeryAI using your WERYAI_API_KEY. Before installing/using: (1) review scripts/video_gen.js yourself (it will be executed locally) and confirm the two API hosts match your expectations; (2) use an API key with limited scope/credits if possible and never commit the key in source; (3) run in an isolated environment if you are concerned about untrusted code; (4) avoid sending private PII or proprietary images to the external service; (5) confirm WeryAI’s terms/privacy and check credit usage and rate limits because each generation consumes credits. The skill source/homepage is unknown — that increases supply-chain risk; if you need stronger assurance, ask the publisher for provenance or use an official client from a known vendor.

Hidden truth reveal (polished → brutal → one line)

Template (seconds below assume duration 10; rescale for 5 or 15): 0–3s hook: what you use daily comes from here (user’s X); 3–8s stark contrast montage (sleek product ↔ labor, factory, waste, or source environment—non-gore, suggestive not graphic); 8–10s one-line summary that lands the thesis. English subtitles with explicit second windows. Default duration is 10 when unspecified.

Dependencies: WERYAI_API_KEY + Node.js 18+. {baseDir}/scripts/video_gen.js + WERYAI_VIDEO_API.md under {baseDir}/resources/. No other Cursor skills required.

Prerequisites

• WERYAI_API_KEY must be set; Node 18+; https images only.
• Non-empty model in JSON; pick from Recommended models or frozen tables.
• Each wait consumes credits.

Security, secrets, and API hosts

• WERYAI_API_KEY: Secret; requires.env / primaryEnv—never commit values.
• API hosts: video_gen.js uses https://api.weryai.com (tasks) and https://api-growth-agent.weryai.com (models list).
• Run in an isolated environment when possible.

Prompt expansion (mandatory)

• Duration: Default 10; 5 or 15 when the user asks and the model allows.
• Timeline scaling: Rescale 0–3 / 3–5 / 5–7 / 7–10 proportionally when duration ≠ 10; keep the four-beat order.

Locked caption rhythm (example for duration 10):

• 0.0–3.0s — Hook naming the daily X (user-supplied): e.g. YOUR DAILY X COMES FROM HERE (keep ≤6 words per line).
• 3.0–5.0s — Polished beat label: e.g. SHELF PERFECT.
• 5.0–7.0s — Harsh beat label: e.g. HIDDEN COST.
• 7.0–10.0s — One-line verdict (user-themed): e.g. THAT’S THE TRUTH (customize to topic).

Visuals: Hard cuts; high contrast lighting; hands / machines / environments—avoid graphic gore or identifiable real people in exploitative frames. Motion every beat.

Typography: Bold sans, white + black outline, lower third, fades; English only on screen unless user opts out.

Audio: Tension bed / impact whoosh if generate_audio true—no named tracks.

Negatives (Kling): illegible captions, watermark, muddy text.

Exception: User supplies finished long prompt + no rewrite—still show full text in confirmation.

Confirmation: Full expanded prompt in the table.

Niche checklist

• Contrast must read in silent skim—image does half the work.
• Middle beats are two emotional temperatures, not a lecture.

Examples below are targets only.

Workflow

1. Text and/or image-to-video (multi-image only SEEDANCE_2_0, ≤3 URLs).
2. Collect product or topic X, optional reference image, tier.
3. Expand with locked windows + visuals; translate user Chinese brief to English on-screen lines unless they request another language.
4. Validate fields vs frozen tables.
5. Confirmation table + full prompt → user confirm.
6. node {baseDir}/scripts/video_gen.js wait --json '…'.
7. Return [Video](url); no code fences around links.

CLI reference

node {baseDir}/scripts/video_gen.js wait --json '{"model":"KLING_V3_0_PRO","prompt":"…","duration":10,"aspect_ratio":"9:16","generate_audio":true,"negative_prompt":"illegible text, watermark, garbled captions"}'
node {baseDir}/scripts/video_gen.js wait --json '{"model":"SEEDANCE_2_0","prompt":"…","duration":10,"aspect_ratio":"9:16","resolution":"720p","generate_audio":true}'

Full reference: WERYAI_VIDEO_API.md.

Definition of done

Video or clear error. prompt contains timed English subtitle block matching the four-beat order, with second ranges scaled to duration (or user’s approved variant).

Boundaries (out of scope)

• Not a journalism fact-check; opinion / hook tone—user responsible for claims.
• No gore; no harassment of real individuals.
• Captions only via prompt; no NLE export spec.
• Use resources/WERYAI_VIDEO_API.md for CLI; no negative_prompt on Seedance.
• Multi-image only SEEDANCE_2_0, max 3.

Example prompts

• Coffee cup: shelf glamour vs harvest-labor montage, verdict line about price to growers
• Smartphone: polished device vs mine/e-waste suggestion, one-line kicker
• Image URL of sneakers → same four-window subtitle schedule
• Fast fashion haul vs factory floor suggestion—verdict about true cost

---

Model and API constraints (frozen for this skill)

Text-to-video

model_key	durations	aspect_ratios	resolutions	audio	negative_prompt	prompt limit
KLING_V3_0_PRO	5, 10, 15	9:16, 1:1, 16:9	(omit resolution)	yes	yes	2000
KLING_V3_0_STA	5, 10, 15	9:16, 1:1, 16:9	(omit resolution)	yes	yes	2000
SEEDANCE_2_0	5, 10, 15	9:16, 1:1, 16:9	480p, 720p	yes	no	2000

Image-to-video

model_key	durations	aspect_ratios	resolutions	audio	negative_prompt	image slots
KLING_V3_0_PRO	5, 10, 15	9:16, 16:9, 1:1	(omit)	yes	yes	single image
KLING_V3_0_STA	5, 10, 15	9:16, 16:9, 1:1	(omit)	yes	yes	single image
SEEDANCE_2_0	5, 10, 15	9:16, 1:1, 16:9	480p, 720p	yes	no	up to 3 images

Recommended models

Tier	model_key	Default duration
⭐ Best	KLING_V3_0_PRO	10
👍 Good	KLING_V3_0_STA	10
⚡ Fast	SEEDANCE_2_0	10 (resolution: 720p)

Default parameters

Field	Value
aspect_ratio	9:16
duration	10
generate_audio	true
resolution	720p for Seedance only
negative_prompt	Kling: illegible text, garbled captions, watermark

Text-to-video (primary)

Collect X → expand four-window captions → confirm → wait.

Image-to-video (optional)

Product still → motion + same schedule.

Tips

• If claims are sensitive, soften to “what they don’t show” rhetoric.
• Verdict line should fit one breath.

Packaged as hidden-truth-reveal-video-gen.