influencer-vetting

Files (1)

• SKILL.md

SKILL.md (first lines)

# Influencer Vetting & Selection

## What This Skill Does

Takes a campaign brief from the user, sources or evaluates creator candidates, scores them
on observable content signals, and produces ranked scorecards with adapted briefs for top
picks. The scoring is transparent — the user sees every weight and every piece of evidence
that drove each score.

The core insight: most influencer vetting is either pure vibes or pure metrics. This skill
does neither. It reads a creator's actual content — the way they tell stories, the hooks
they use, the emotional register they operate in, how they integrate products, what their
audience conversations look like — and assesses fit against the specific campaign, not
against generic benchmarks.

## What You Can and Cannot Assess

This is the foundation. Be honest about it throughout.

**What content intelligence shows you:**
- How a creator communicates — their hooks, narrative style, emotional tone, production
  approach, storytelling patterns
- What topics and themes they consistently cover
- How they integrate products and brands into content (naturally vs. forced)
- Creative range — do they have one trick or many?
- Posting consistency and content volume
- Engagement lift against the creator's own baseline as a proxy for audience quality
  (not audience identity)
- What their audience says back in comments — the texture of the community
- How their content compares to the broader landscape in their category

**What you cannot assess from content alone:**
- True audience demographics (age, gender, location, income)
- Fake follower percentages or bot activity
- Conversion rates from past partnerships
- Pricing and rate cards
- Professionalism, responsiveness, contract reliability
- Whether high engagement translates to purchase behavior

**What this means for output:**
Every scorecard includes a "Verify Externally" section that flags exactly what the user
needs to check through other means — audience quality tools (HypeAuditor, etc.), pricing
negotiations, reference checks with past brand partners. Don't pretend the analysis is
complete. It's the content intelligence layer — the deepest, hardest-to-get layer — but
it's one layer.

## Phase 1: Campaign Brief Intake

No analysis without context. Creator fit is entirely relative to the campaign. A creator
who's perfect for awareness is wrong for conversion. A creator who's great for Gen Z
lifestyle is wrong for B2B decision-makers.

**Before doing anything, get these from the user:**

**Required:**
- **Brand name** — Who is this for?
- **Campaign goal** — What does success look like? (Awareness, consideration, conversion,
  content library, community building)
- **Target audience** — Who are we trying to reach? Be specific — not just "millennials"
  but "health-conscious women 25-35 who are curious about functional beverages but skeptical
  of wellness hype"
- **Platform focus** — Where does this need to live? (TikTok, Instagram, YouTube, LinkedIn,
  multi-platform)

**Strongly recommended:**
- **Content format needs** — What kind of content? (Short-form video, long-form review,
  lifestyle integration, educational, UGC-style, polished production)
- **Brand voice/positioning** — How does the brand talk? What's the vibe? (Premium,
  approachable, scientific, playful, rebellious, etc.)
- **Existing brief or key messages** — If they have a creative brief, campaign messaging,
  or key talking points, get them. This is what Phase 5 adapts.
- **Budget tier** — Not exact numbers, but: nano (<10K followers), micro (10K-100K),
  macro (100K-1M), mega (1M+)? This shapes who to source.
- **Category/vertical** — What space is this in? (Beauty, fitness, tech, food, finance, etc.)

**Optional but valuable:**
- **Competitors to avoid** — Creators who've recently worked with these brands are flagged
- **Must-haves or dealbreakers** — Anything specific? (Must be based in US, must do
  long-form, can't have worked with Brand X in last 6 months)
- **Number of creators needed** — How many final picks?

**If the user just drops handles without a brief**, ask for the brief. You can't score
fit without knowing what they're fitting against. Be friendly about it: "Before I vet
these, I need to understand the campaign so I can score them against something real.
What's the goal, who's the audience, and what kind of content do you need?"

## Phase 2: Creator Sourcing

Two entry paths. The user either gives you a list to vet, or asks you to find candidates.

### Path A: User Provides a List

The user drops handles (e.g., "@creator1, @creator2, @creator3"). Proceed directly to
Phase 3 with their list.

### Path B: Discovery Mode

When the user wants you to find candidates, use **web search at scale** as the sourcing
engine. The goal is a longlist of **30-50 candidates** — cast a wide net, then let the
content analysis narrow it down.

**1. Web search at scale (run in parallel)**
Launch 4-6 parallel web search agents simultaneously, each targeting a different angle:
- "[category] influencers [platform] 2026"
- "best [niche] creators on [platform]"
- "[brand vertical] content creators to watch"
- "[category] TikTok/Instagram/YouTube creators brand partnerships"
- "corporate/office/workplace [platform] creators" (for B2B campaigns)
- Adjacent category searches (e.g., for fintech: corporate humor, accounting, entrepreneurship)
- Competitor brand partnership searches (who has worked with similar brands?)

Each parallel agent should return 10-15 candidates with handles. The model's training data
also contains substantial knowledge about prominent creators across categories — use it.
Think about who you know in the space. The goal is raw volume: 30-50 names with handles
before analysis begins.

**2. Resolve names to real handles**
A name without a verified handle is a dead end — the wrong handle silently tracks the
wrong account. `lookup_brands` resolves a name against Adology's central directory and

From adologyai/content-intelligence-plugin, Apache-2.0.

More from adologyai/content-intelligence-plugin