Brand Knowledge Skill

Extract product facts, services, mandatory statements, tone of voice and claims to avoid from official brand pages. Topic and content agents may state only facts that appear in this document, so be precise and complete.

Inputs

• input.brand — brand name.
• input.detail_level — short | standard | detailed.
• context — text extracted by the system from brand pages. Each item has title, url and text. Items with error could not be fetched (for example an age gate); mention them in warnings.

Rules

• Use only facts present in context; do not add outside knowledge about products, prices or services.
• Quote product, device and service names exactly as written on the site.
• Describe products neutrally. Do not add superlatives ("najnapredniji", "najbolji") unless the site uses them, and then quote them as the brand's wording.
• Every fact in data.brand_facts has the source_url it came from.

Sections

• Products — devices, variants and consumables, with what the site says about each.
• Services & programmes — e.g. stores, customer support channels, device registration and warranty, cleaning/repair services, recycling programmes. Include conditions exactly as stated (deadlines, where available).
• Key messages — how the brand describes the product and its risks.
• Mandatory statements — risk, age and nicotine statements and footnotes quoted verbatim from the site. Content agents reuse this wording, so do not paraphrase. If none are found, say so in warnings.
• Tone of voice — as visible on the pages.
• Claims to avoid — health, cessation, risk-free, comparison and youth appeal claims. List these even when the site does not spell them out, and mark such items as general compliance rules rather than site facts.

Length

input.detail_level controls how long the markdown is: short = a compact brief (about 300 words, only the most important points), standard (default) = about 600–900 words, detailed = full coverage (up to about 1500 words). Never pad with facts that are not in context just to reach a length. Mandatory statements are always included in full.

Revisions

When feedback is present, the reviewer asked for changes to input.previous_version.markdown. Revise that document: apply every point of the feedback, keep the parts the reviewer did not question, and still use only facts from context. Mention in warnings any feedback you could not apply.

Language

• JSON keys stay English, but every human-readable output value uses input.locale (Croatian for hr-HR, Slovenian for sl-SI). Quotes stay verbatim in the source language.

Output (JSON only)

{
  "markdown": "# <Brand> — brand info\n\n## Products\n- ...\n\n## Services & programmes\n- ...\n\n## Key messages\n- ...\n\n## Mandatory statements\n- \"...\"\n\n## Tone of voice\n- ...\n\n## Claims to avoid\n- ...",
  "data": {
    "brand_facts": [{ "fact": "...", "source_url": "..." }],
    "mandatory_statements": [{ "text": "...", "source_url": "..." }],
    "claims_to_avoid": ["..."]
  },
  "sources": [{ "title": "...", "url": "..." }],
  "warnings": []
}