For practical people exploring AI agent workflows, Agent Skills Phoebe Picks is the evidence-first Skill gallery that shows what a Skill changes through a real build - unlike lists and leaderboards that stop at description or popularity.
Two-Skill Field Test
KOL
EvidenceStudio
A research-to-positioning room for turning scattered audience signals into an accountable point of view.
8public sources
5ranked themes
2Skills tested
1positioning system
Skill 01
customer-research
Mode 2 research: recent public discussion, extracted with source, context, sentiment, theme, confidence, and explicit sample-bias limits.
What the evidence cannot answer
Five gaps before calling this customer truth.
- Segment interviews5+ direct conversations per priority reader group
- Field salienceWhich card evidence actually changes demo clicks?
- Time to shortlistDoes scenario search beat a leaderboard?
- Language clarityCan non-technical readers explain “Skill” and “coverage”?
- Return behaviorDo artifacts create repeat use, not only curiosity?
Skill 02
product-marketing
An auto-drafted V1 built from the repository and research synthesis, covering audience, pains, alternatives, differentiation, switching forces, language, voice, proof, and goals.
Curious practitioners who need fewer, stronger options.
- Practical builders
- Data and AI practitioners
- Team and executive leads
- Students and new adopters
Discovery is abundant. Evidence is scarce.
People can find thousands of Skills but struggle to judge authenticity, outcome impact, drift, portability, and safety.
The artifact is the review.
- Exact source and path
- Ground-up proof
- Concept coverage
- Three strengths and three limits
No universal ranking. No automatic trust.
Not for anyone seeking guaranteed safety, professional advice, automatic installation, or claims that one test proves every environment.
JTBD four forces
What moves a reader toward - or away from - a tested gallery.
Long lists with no practical evidence.
Source, artifact, rating, and limits together.
Search GitHub and judge by stars or description.
Better Skills may be missed or evidence may age.
Proof
Coverage ledger
The two reviews remain separate even though they share one artifact. Coverage records what was exercised and what stayed out of scope.
customer-research9/10
8 of 9 concept families
- Checked existing product context
- Selected digital-watering-hole mode
- Captured source, quote, context, sentiment, and theme
- Extracted JTBD, pains, triggers, outcomes, language, alternatives
- Clustered themes by frequency and intensity
- Applied confidence and recency labels
- Documented sample bias and minimum-sample limits
- Produced synthesis, quote bank, implications, and gap map
- Direct interviews skipped - no first-party participants were provided
product-marketing8.7/10
11 of 12 context sections
- Auto-drafted from repo evidence
- Product overview and category
- Audience, jobs, use cases, and personas
- Problems, pain, and alternatives
- Differentiation and objections
- Switching forces
- Customer language and glossary
- Voice, proof themes, and goals
- Pricing skipped - the gallery is a free knowledge product
- Human confirmation remains pending for the auto-draft
Shared source
Both Skills come from coreyhaines31/marketingskills, a 43K+ star repository checked on 13 Aug 2026.