L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Design et création · SKILL.md

iopho-product-context

iopho-product-context

Generate structured product context files for video production projects. Runs an interactive intake questionnaire (or auto-detects from existing docs), outputs a context.md that all other iopho skills read. Use when starting a new video project, onboarding a client, or when any iopho skill needs product/audience/brand information. Triggers: "product context…

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: iopho-product-context description: > Generate structured product context files for video production projects. Runs an interactive intake questionnaire (or auto-detects from existing docs), outputs a context.md that all other iopho skills read. Use when starting a new video project, onboarding a client, or when any iopho skill needs product/audience/brand information. Triggers: "product context", "video project setup", "new project", "context file", "onboarding questionnaire", "client intake". user-invocable: true argument-hint: "[project-dir] [--quick] [--from-file README.md] [--lang en|zh|mixed]" updated: "2026-03-19"

iopho-product-context

Generate structured context.md for video production — the foundation file all other iopho skills read.

When to Use

  • Starting a new video project (product demo, launch video, explainer, VSL)
  • Onboarding a client project (like a client project)
  • Any iopho skill asks for context but none exists yet
  • Updating an existing context file with new information

Quick Start

/iopho-product-context ./exp/my-project
/iopho-product-context ./exp/my-project --quick           # 5 core questions only
/iopho-product-context ./exp/my-project --from-file README.md  # pre-fill from docs

How It Works

Mode 1: Interactive (default)

Ask the user these questions IN ORDER. Accept answers in any language (CN/EN/mixed).

Core Questions (always ask):

  1. What's the product? — Name, URL, one-sentence description
  2. Who is it for? — Primary audience + their pain point
  3. What makes it different? — Top 3 differentiators from competitors
  4. What video do you want? — Type (demo/explainer/launch/teaser), duration, platforms
  5. Any brand rules? — Colors, fonts, tone of voice, design system files

Deep-Dive Questions (ask unless --quick):

  1. Who are the competitors? — List 2-5 competitors, their strengths/weaknesses
  2. Founder visibility? — Should founders/team appear? (public / anonymous / team-only)
  3. Hero feature? — The ONE feature that must be the centerpiece of the video
  4. What to avoid? — Anything off-limits (features, messaging, visuals)
  5. Reference videos? — Any video styles they like (provide URLs → feed to iopho-searching-videos)
  6. Languages? — Which languages for VO and subtitles
  7. Call to action? — What should viewers do after watching

Mode 2: Auto-Detect (--from-file)

Before asking questions, scan the project directory for existing context:

Look for and read (if present):
├── README.md              → product name, description, features
├── package.json           → name, description, keywords
├── **/brand.ts            → colors, fonts (Remotion projects)
├── **/*DesignPrompts*     → design system (client project pattern)
├── **/questionaire.md     → existing intake (client project pattern)
├── **/context.md          → previous context file (update mode)
└── website URL            → scrape tagline, features, pricing (if provided)

Pre-fill answers from detected files, then ask user to CONFIRM or CORRECT each field. Skip questions already answered by auto-detection.

Mode 3: Update

If context.md already exists in the project directory, read it and ask:

"Found existing context.md. What needs updating?"

Only modify the fields the user mentions. Preserve everything else.

Output

Generate {project-dir}/context.md using the template at templates/context-template.md.

Output rules:

  • YAML frontmatter: machine-readable fields (name, URL, colors as hex, platforms as array)
  • Markdown body: human-readable prose (features, audience, competitors, notes)
  • All hex colors validated (6-char format)
  • All URLs as full https:// format
  • Bilingual content preserved as-is (don't translate user's words)

Validation

After generating context.md, verify:

  • Product name is not empty
  • At least 1 target audience defined
  • At least 1 video platform specified
  • At least 1 language specified
  • Brand colors are valid hex (if provided)
  • Founder visibility preference is set

If any critical field is missing, ask the user — don't guess.

Real-World Examples

[App Name] (a reader/learning app):

  • Product: AI read-later app → Explainer video, 78s, YouTube/Product Hunt
  • Brand: #007AFF primary, SF Pro font, warm/approachable tone
  • Hero feature: AI reflow + instant summary
  • Context existed in: {project}/video-strategy.md

[CLI Tool] (a developer networking tool):

  • Product: P2P agent collaboration → Launch video, 90s, YouTube/GitHub
  • Brand: #1D3557 deep ocean, #E63946 lobster red, cinematic/authoritative
  • Hero feature: ASCII globe + npx install
  • Context existed in: {project}/questionaire.md + {project}/product-codex.md
  • Founder visibility: anonymous (user requested)

Related Skills

  • /iopho-video-director — calls this skill as Phase 0 step 1
  • /iopho-searching-videos — uses reference video URLs from context
  • /vsl-storyboard-writer — reads context for audience/positioning
  • /storyboard — reads context for product understanding
Skills enregistrées