S
Scores any cited source for credibility, freshness, authorship and bias. Flags marketing-disguised content. Used before any claim ships.
Used by Content team · Market Expert
Platform · Capability deep-dive
The agent's SOPs. Pre-built by senior practitioners.
A skill is a plug-in chunk of expertise: a method + the tools it needs + the credentials to use them. Enable one and the agent gains a discipline a 5-year specialist would bring on day one. Author your own once and apply to any agent.
Built by people who've done the workEnable in seconds, ship the same dayScoped tools, scoped credentialsAuthor your own · share across teams
The wedge
A skill is not a prompt
The difference between "tell the AI to do it" and a real method. The first is a hope; the second is what wins repeatably.
A prompt
"Vet these sources for me."
You wrote a clever instruction. The model interprets it differently each time. No tools attached, no quality bar.
- Reinvented every conversation
- Inconsistent — sometimes good, sometimes not
- No access to the tools the method actually needs
- No one to call when it gets it wrong
A skill
Source Vetting — the discipline.
A method written by someone who's done it for a decade. Tools bundled, credentials handled, scored against ground truth. Repeats the same way every time.
- Versioned method — one source of truth
- Consistent quality, scored continuously
- Bundled with the tools the method needs
- Maintained — updates as practice evolves
The catalog
Real skills. Real names.
The catalog grows weekly. Here is the current shelf — with the agents that use each one. Author your own to extend.
AllResearchEditorialMethodTechnicalOperationsCustom18 skills shown
H
Compares an idea or product claim against actual evidence. Down-weights buzzwords, surfaces what's real. Stops AI slop at the door.
Used by Critique · Content Writer
K
Maintains a structured graph of entities, relationships and claims. Detects drift, proposes corrections, keeps facts queryable.
Used by Knowledge Graph Agent
B
Holds your brand voice as a model: banned phrases, signature constructions, tone rules. Every draft scored before sign-off.
Used by Critique · Content Team Lead
W
Multi-source research with citation gathering and provenance tracking. Bundled with a real SERP API and web fetcher — not the model's memory.
Used by Market Expert · Idea Generator
M
Breaks a fuzzy goal into a sequenced plan of subtasks each agent can execute. Coordinates parallel work and rolls results back up.
Used by Strategy Lead · Team Leads
X
Reads, transforms and writes spreadsheets with formula awareness. Pivot tables, lookups, charts — all done programmatically.
Used by Data Analyst · Reporting Agent
C
Traces every AI citation back to its source. Identifies which third-party pages move which answers. The backbone of attribution.
Used by Data Analyst · Competitor Analyst
P
Builds advanced buyer personas per product: who buys, why, criteria, objections, language. Validated, queryable, updated.
Used by Company Expert
Sd
Generates valid Schema.org markup (Organization, Product, FAQ, Comparison, Article) for any page. Tested against Google Rich Results.
Used by Owned-Site Technical
Cr
Tests every AI crawler (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) against every public route. Flags blocks, render fails, redirects.
Used by Owned-Site Technical
O
Drafts targeted outreach to specific publications and reviewers in your citation graph. Real reasons, no templates. Your sign-off before send.
Used by Authority Outreach
A
Restructures existing pages so the answer sits at the top, claims become quotable sentences. Same content, AI-citable shape.
Used by Content Writer
I
Continuous verification of every connector: auth, rate limits, schema, latency. Auto-remediates within minutes of a breakage.
Used by Technical Controller
Q
Samples each agent's output, scores against ground truth and consuming-team feedback. Trends scores over time.
Used by Performance Monitor
D
Statistical alerts on output distribution. Catches degradation before it surfaces in your dashboard.
Used by Performance Monitor
Cb
Your in-house example: flags anything in a draft that needs your legal team's sign-off. Authored by you, applied to every Writer.
Used by Your custom skill
R
Turns multi-agent work into a human-readable summary: what happened, who did what, what to decide next.
Used by Reporting Agent · Team Leads
Anatomy of a skill
What's actually inside one
A worked example: Source Vetting, the skill that decides whether a citation belongs in a draft.
S
Research skill
Source Vetting
Scores any cited source for credibility, freshness, authorship and bias. Runs before any claim ships.
The method
A written procedure: score each source on credibility, freshness, authorship and bias; flag marketing-disguised content; require a passing threshold before a citation can enter a draft.
Bundled tools
Web fetcherCitation crawlerArchive checkerSERP API
Credentials & scope
Read-only tokens scoped to this skill alone. Expiring, revocable, never shared across agents.
Quality & assignment
Scored continuously against ground truth. Assigned to Content Writer, Critique and Market Expert.
Tailor & extend
Your in-house SOPs, captured once
Three steps to your own skill. Write it once; every agent that needs it gains the discipline.
i
Write the method
Name, description, the actual method. Plain prose — the discipline as you do it, not as a generic AI assistant would imagine it.
name: Compliance Review description: Flags anything that needs legal sign-off. method: Check unsupported claims, regulated keywords, comparatives…
ii
Allowlist the tools
Pick the tools the method needs. The skill gets exactly those — no more. Credentials scoped to that skill alone.
tools: - web_fetch # read-only - citation_crawler - internal_kb_query credentials: scoped, expiring
iii
Assign to agents
Apply the skill to one agent, a team, or every agent that needs the discipline. Same skill, multiple workers — one source of truth.
assigned_to: - Content Writer - Marketing Writer (cloned) - Sales Drafter # propagates automatically
Your AI workforce for AI search
See a skill working on your business
30-minute working session. Pick one skill, watch it run against your data — or describe one of your own SOPs and we'll show how it would land.
Not ready to talk? Get a free AI-visibility snapshot →