エージェントライブラリ

Turn a bare website URL into a full batch of on-brand Meta ad creatives, statics and video.

Inspect and operate Meta Ads Manager through Playwriter.

Turn approved Meta advertising hypotheses into grounded production briefs, visual prompts, copy variants, asset batches, and quality reviewed handoffs.

Design and analyze controlled Meta ad creative tests across angles, hooks, formats, offers, and audiences.

Analyze Meta advertising performance across account structure, creative, delivery, audience, funnel, and business economics.

Audit Meta advertising measurement from the browser event through Events Manager, attribution, deduplication, and reporting.

Review and draft Google Ads headlines, descriptions, assets, and landing page recommendations.

Analyze Google Ads performance across campaigns, queries, keywords, ads, assets, landing pages, bidding, conversion goals, and business economics.

Inspect and operate Google Ads through Playwriter.

Review Google Ads search terms, identify irrelevant or harmful query patterns, and prepare a safe negative keyword plan.

Audit a site, research and cluster keywords, inspect search intent, build briefs, improve internal links, and create an evidence based SEO plan.

Audit and improve internal linking with verified page relationships, useful anchors, crawl paths, and measurable implementation plans.

Build a source backed keyword universe, validate search intent, cluster related queries, detect cannibalization, and prioritize page opportunities.

Evaluate external link and partnership opportunities using current evidence, editorial relevance, audience fit, risk, transparency, and expected effort.

Run an evidence based SEO production pipeline from raw data through current search result validation, clustering, briefing, drafting, quality assurance, internal links, human publishing approval, and...

Turn a validated keyword cluster and current search results into a factual, differentiated content brief.

Audit crawling, indexation, rendering, canonicalization, redirects, structured data, site architecture, performance, and technical search risks.

Use Engain to research relevant Reddit opportunities, maintain an approval queue, verify community rules and claims, and audit manually published replies against live permalinks.

Draft useful transparent Reddit replies from the full thread, community rules, verified expertise, and approved product claims.

Research buyer questions on Reddit, map community rules, identify relevant discussions, draft transparent helpful replies, and monitor legitimate brand participation.

Map Reddit communities, mine buyer language, find current discussions, and score legitimate participation opportunities.

Build grounded email campaign calendars from business events, product priorities, audience segments, and deliverability constraints.

Audit permission, authentication, list health, sending patterns, content risk, suppression, and inbox performance for lifecycle email programs.

Design and prepare verified Klaviyo lifecycle flows as drafts.

Plan, build, audit, and improve Klaviyo lifecycle marketing.

Draft a short cold email sequence from verified audience, offer, proof, contact source, and jurisdiction inputs.

Audit, validate, deduplicate, segment, and document business outreach lists before sending.

Audit outreach infrastructure, sending readiness, domain authentication, mailbox setup, list governance, suppression handling, and approval controls.

Classify outreach replies, preserve opt outs, draft grounded responses, and route opportunities.

Design, prepare, audit, and improve responsible business outreach in SmartLead.

Prepare creator briefs and review deliverables against approved terms, claims, brand rules, usage rights, and platform requirements.

Build verified creator shortlists from public evidence and supplied campaign criteria.

Run a creator partnership end to end.

Evaluate creator content performance, attribution, brand safety, usage rights, disclosure status, and paid amplification readiness.

Review the exact current browser state before a consequential Playwriter action and require explicit approval that matches the target and effect.

Diagnose failed Playwriter browser workflows using visible state, relevant browser errors, minimal sanitized network facts, and a bounded recovery plan.

Operate an authorized logged in browser through Playwriter with target confirmation, current page inspection, bounded actions, approval gates, and destination readback.

Coordinate approved Playwriter workflows across paid media, SEO and CMS, email marketing, Reddit research, reporting, and client operations while preserving channel specific approval boundaries.

Extract complete authorized dashboard data through Playwriter and verify scope, pagination, downloads, counts, and representative records before analysis.

Perform approved browser uploads and downloads through Playwriter while proving file identity, destination, scope, validation state, and final draft status.

Run N Playwriter agents concurrently on one browser without them colliding.

Capture a work session in an Obsidian daily note and promote durable decisions, outputs, failures, and next actions to their owner notes.

Build and run durable Obsidian memory for campaigns, SEO, email, content, reporting, and client learning loops.

Create resume ready Obsidian handoff packets and coordinate disjoint agent memory work with one integration owner.

Build and operate an Obsidian vault as durable memory for AI agents.

Perform exact official Obsidian CLI reads, creates, appends, searches, link audits, property checks, daily note operations, and history inspection with vault targeting and readback.

Create and maintain resume ready Obsidian project notes and decision ledgers with current state, evidence, artifacts, risks, validation, and next actions.

Index and retrieve approved Markdown memory with QMD keyword, vector, and hybrid search.

Design a small predictable Obsidian vault or client root with folder purposes, Brain, External References, source ownership, and privacy zones.

Audit Obsidian vault health across stale context, source ownership, unresolved links, orphans, dead ends, QMD retrieval, privacy boundaries, sync, and backup evidence.

Classify vault content, enforce personal and per client storage and retrieval boundaries, and review public memory packages for private data.

Diagnose missing, stale, contradictory, mislinked, or incorrectly indexed Obsidian memory and prepare evidence based recovery from local history, Sync history, or backup.

Build an evidence based audience profile from customer language, behavior, pains, desired outcomes, alternatives, objections, and buying triggers.

Model acquisition and retention economics, compare marketing channels, and set evidence based test budgets and stop rules.

Draft and review grounded marketing messages from an approved offer, audience evidence, proof, and channel context.

Build a grounded marketing foundation from an offer, audience evidence, unit economics, and proof.

Create marketing copy from a student owned proof bank.

Create short teaching scripts from verified proof and a chosen marketing pillar.

Build or refresh minimal sourced context packets for an AI agent from Brain, current daily notes, project notes, context, and exact evidence.

Design independent review agents, approval gates, and stop conditions for risky workflows.

Initialize persistent project context for long running agent work.

Turn a repeated marketing workflow into a concise installable skill.

Turn loose ideas into precise, grounded, executable prompts for AI agents.

Design bounded multi agent workflows with clear ownership and validation.

Review analytical conclusions and confidence against the available evidence.

Independently test implemented behavior across realistic paths.

Review customer facing copy against approved evidence and policy.

Review a proposed destructive or difficult to reverse action before execution.

Run the repository checks required before code handoff or publication.

Audit research claims against current primary sources and supplied evidence.

Verify the visible browser target and proposed action before a consequential click.

Test completed work against the user defined finish line and return evidence.

Mirror your Claude Code and Codex session transcripts into your Obsidian vault as searchable notes, then keep them current with a background job.

Turn a substantial project into a supervised agent workflow with bounded delegation, guardrails, integration, and final verification.

Plan and produce grounded branded creative batches with Claude and an approved image generation tool such as Higgsfield.

Implement one approved and bounded work package with tests and evidence.

Convert approved research and requirements into a precise implementation plan.

Turn a failed attempt into a durable project constraint.

Research and plan complex agent projects without implementing them.

Record verified project state for future agents and sessions.

Iterate on a bounded task by repeating the same objective with accumulated evidence until a truthful completion promise is satisfied.