brand-builder
GitHub指导构建知识库集，涵盖品牌发现、信用检查、数据抓取及缺口填充。用于创建KS、添加品牌/红人/搜索词或监控竞品。
Trigger Scenarios
Install
npx skills add NeverSight/learn-skills.dev --skill brand-builder -g -y
SKILL.md
Frontmatter
{
"name": "brand-builder",
"description": "Guides knowledge set construction including feed management and brand discovery. Use when creating knowledge sets, adding or removing brands, influencers, search terms, or discussion feeds. Triggers on: \"create knowledge set\", \"add brand\", \"add influencer\", \"track competitor\", \"monitor\", \"build a KS\"."
}
Building Knowledge Sets
Step 1: Discovery
Start by running discover_brands to find brands by name, category, or URL. This gives you a picture of the competitive landscape before building anything.
Step 2: Credit Awareness
After discovery, call whoami to check the user's available credits. Before calling trigger_fetch, estimate the cost based on the number of feeds being fetched. Inform the user of the expected credit usage before proceeding. Never trigger a fetch without the user understanding the cost.
Step 3: Create and Populate the Knowledge Set
Use create_knowledge_set with a descriptive name ("DTC Skincare Competitors", "Q2 Campaign Benchmarks", "Protein Brand Landscape").
When adding 3+ feeds, use batch_add_feeds instead of individual add_feed calls. This is faster and reduces round-trips. Only use individual add_feed for single additions or when you need to handle each feed's response separately.
Conditional Logic by Input Type
User gives a brand name:
- Run
discover_brandsto find it and resolve platform handles - Check
list_knowledge_setsto see if the brand already exists in another KS - If found elsewhere, inform the user before duplicating
User gives a URL:
- Extract the platform and handle from the URL
- Add directly via
add_feed-- no discovery step needed
User gives a category (e.g., "energy drinks"):
- Run
discover_brandsby category to get a list of brands - Present the options and let the user pick
- Suggest related brands they may not have considered
Step 4: Trigger Data Collection
Call trigger_fetch to start scraping. This runs asynchronously via Temporal workflows. You can use the feedNames parameter to selectively refresh specific feeds instead of the entire Knowledge Set.
Check progress with get_workflow_status. Scraping typically takes a few minutes per feed depending on content volume.
Step 5: Get Suggestions and Fill Gaps
After feeds are added, call get_suggestions to identify gaps in the Knowledge Set:
- Missing platforms for existing brands (e.g., brand has Instagram but not TikTok)
- Missing feed types (e.g., no search terms, no discussion feeds)
- Related brands the user might want to track
Review the suggestions with the user and add recommended feeds using batch_add_feeds.
Complete Onboarding Flow
The full sequence:
- Discover --
discover_brands - Check credits --
whoami - Create KS --
create_knowledge_set - Add feeds --
batch_add_feedswith all selected brands, creators, search terms, discussions - Fetch data --
trigger_fetch(after confirming credit cost) - Monitor --
get_workflow_statusuntil complete - Fill gaps --
get_suggestionsto identify missing coverage - Expand --
batch_add_feedswith suggestions, thentrigger_fetchagain
Feed Types Reference
Brand Feeds
Track paid ads and organic social content. Sources: facebook-ad-library, tiktok-ad-library, instagram-profile, tiktok-profile, youtube-channel, facebook-page, twitter-profile, linkedin-profile, threads-profile.
Influencer Feeds
Track a specific creator's content. Sources: instagram-profile, tiktok-profile, youtube-channel, twitter-profile, threads-profile.
Search Feeds
Monitor keywords across platforms. Sources: tiktok-search, instagram-search, youtube-search.
Discussion Feeds
Track Reddit subreddits. Provide the subreddit name without the r/ prefix.
Error Recovery
Brand not found in discovery DB: Ask the user for the brand's social media handles directly. You can add feeds with explicit handles even if the brand is not in the discovery database.
Creator not in discovery database:
Creator discovery is not available via MCP. Ask the user for the creator's handle directly and add via add_feed with feedType: 'influencer'.
add_feed with invalid handle: The tool will return an error. Inform the user the handle could not be resolved. Ask them to verify the handle exists on the platform, check for typos, and try again.
Tips
- Add at least 2-3 competitors for meaningful competitive benchmarking
- Search terms work best when specific ("collagen supplement" > "health")
- Subreddit monitoring captures authentic consumer voice that brand content misses
- You can update a Knowledge Set at any time to add or remove feeds
Version History
- e0220ca Current 2026-07-05 23:27
