Claude Agents for B2B, GTM, and RevOps: The Complete Playbook

Sources: Claude Agents for GTM (henriquecaner/next-level-outreach, 40 skills) · Claude Agents for B2B (5 agents: lead qualification, account research, onboarding, competitive intelligence, proposal generation) · Claude Agents for B2B SaaS: 50+ Agents (gtmagents, alirezarezvani, VoltAgent, coreyhaines31, sales-skills, zubair-trabzada, louisblythe, ComposioHQ, indranilbanerjee, modest-curator478, BehiSecc)

This playbook combines all three agent collections into one end-to-end resource. It covers what agents are and why they matter for complex B2B work, the four GTM agent types (outreach, content, campaign, daily ops), the five B2B deal agents (lead qualification, account research, client onboarding, competitive intelligence, proposal generation), and the complete 50+ agent directory for sales, marketing, product, CRM, customer success, finance, and multi-agent orchestration.

What an agent is and why it is different from a prompt or skill

A prompt is a one-shot request. A skill is a saved workflow that runs once, produces output, and stops. An agent goes further. It can spawn subagents to handle parallel tasks, loop until a condition is met, call external tools via MCP, read and write files, and verify its own output before finishing.

The difference in practice: a skill writes one cold email. An agent scrapes a list, classifies every lead against your ICP, enriches the ones that pass, writes personalised first lines for each, and uploads the result to Instantly, all in one session.

For GTM work, the distinction matters because most of your highest-value workflows are multi-step. A single skill handles one step. An agent handles the whole motion.

For B2B work specifically, the defining characteristic is context volume. A B2B deal might involve 5-10 stakeholders across multiple departments, a 3-6 month sales cycle, technical evaluation requirements, procurement processes, and legal review. The amount of context required to work a deal well is enormous. Agents handle context volume better than humans. A B2B research agent can read 10 sources about a target account, synthesise them into a decision-ready brief, and do it in under 5 minutes. The same task done manually takes 45-60 minutes and is rarely done as thoroughly.

How to install any agent

# Method 1: npx
npx skills add [org]/[repo]

# Method 2: Plugin marketplace
/plugin marketplace add [org]/[repo]
/plugin install [skill-name]

# Method 3: Git clone
git clone https://github.com/[org]/[repo] ~/.claude/skills/[skill-name]

# Method 4: Copy to skills directory
cp -r [skill-folder] ~/.claude/skills/

# Verify installed
npx skills list

Skills load progressively. At session start, the agent sees only each skill name and description (~100 tokens). The full SKILL.md body loads only when the skill is called. A single Claude Code session can host hundreds of skills without bloating the context window.

The nine agent types in this playbook

Agent	What it runs
Outreach agent	Full outbound motion: ICP, lists, copy, LinkedIn, calls, replies. 40 skills.
Content agent	LinkedIn content, morning digest, competitor intel, client reports, AI articles. 7 skills.
Campaign agent	Angles, architecture, workflow orchestrators chaining 8+ agents.
Daily ops agent	Scheduled recurring tasks: digest, pipeline check, inbox triage, campaign review.
Lead qualification agent	AI scoring, automatic routing, n8n workflow, no manual qualification step.
Account research agent	Company overview, stakeholder map, tech stack, buying signals, angle recommendations.
Client onboarding agent	Folder setup, brief population, Linear project, Slack notification. 2 minutes from signed contract.
Competitive intelligence agent	Landscape map, battlecards, win/loss frameworks, scheduled monitoring.
Proposal generation agent	Reads account brief, generates structured proposal in under 5 minutes.

Source: github.com/henriquecaner/next-level-outreach

40 skills across 7 folders. One canonical brief file. The full B2B outbound motion from a single onboarding command.

The outreach-brief.md: the agent's memory

The outreach brief is the canonical context file for the outreach agent. It contains: company name and one-sentence value prop, ICP definition (trigger-based, not demographic), buyer personas by seniority with pain and objection mapping, value propositions ranked by ICP priority, competitor names with differentiation angles, and buying signals with detection methods.

Without it, every skill runs blind. Cold email references one value prop, LinkedIn DM references a different one, call script uses a different ICP. With it, every touch is consistent because every skill reads the same source.

Build the brief by running:

/nl-outreach:onboard

The command interviews you using 8 skills in sequence and writes the complete brief to disk. Takes 5-8 minutes in guided mode. Run once per client engagement. Update when the ICP shifts, a competitor enters, or the offer changes.

The 7 folders and what each handles

strategy/ (4 skills)

icp-definer builds trigger-based ICP definitions, not demographic filters. persona-definer creates contact-level cards by seniority. pain-identifier maps observable signals to specific angles. offer-definer rewrites features as outcome-focused offers. Run these once per client engagement and update when the ICP shifts.

list-building/ (4 skills)

company-finder filters via Apollo, Clay, and lemlist. people-finder targets via Sales Nav and LinkedIn. niche-data-finder surfaces signals from G2 and ProductHunt. list-builder deduplicates and formats to CSV. The output of this folder feeds directly into copywriting.

copywriting/ (5 skills)

copywriting-first-touch writes signal-led emails under 100 words. copywriting-follow-up handles touch 2 and the breakup email. copywriting-vp-sequence adjusts tone for VP-level contacts. copywriting-refiner runs a six-point checklist against any draft. copywriting-analyzer benchmarks against data from 244,000+ campaigns.

outreach/ (3 skills)

linkedin-sequence writes 2-DM maximum sequences using the 2-DM max rule. cold-call-script produces a 6-part call framework with objection handlers. video-outreach connects to Tavus and HeyGen for personalised video outreach.

campaign/ (3 skills)

campaign-angle-finder generates three distinct angles per ICP before writing anything. outbound-campaign-architect designs the full campaign structure. gtm-action-thinker stress-tests any GTM play before execution.

replies-pipeline/ (3 skills)

reply-handler covers 8 reply types with full response scripts for each. outbound-analyst benchmarks campaign metrics against 244,000+ campaign data. pipeline-analysis reviews the pipeline stage by stage and flags stalls.

automation/ (3 skills)

n8n-workflow-builder produces exportable JSON for any described workflow. n8n-debugger covers the 6 most common n8n errors with fixes. website-scraper builds Playwright and BeautifulSoup scripts for any scraping target.

Install and run the outreach agent

# Install via plugin marketplace
/plugin marketplace add henriquecaner/next-level-outreach
/plugin install nl-outreach@nl-outreach-marketplace

# Or git clone for immediate access
git clone https://github.com/henriquecaner/next-level-outreach
cd next-level-outreach && claude --plugin-dir .

# For Cowork: Settings > Customize > Skills > Create skill > Upload skill zip files

# Step 1: Run the onboarding command (once per client or business)
/nl-outreach:onboard

# Step 2: Invoke skills by name
/nl-outreach:copywriting-first-touch
/nl-outreach:campaign-angle-finder
/nl-outreach:linkedin-sequence
/nl-outreach:reply-handler
/nl-outreach:pipeline-analysis

# Or describe in plain language:
"Write a cold email for a VP Sales at Series A SaaS who just posted two SDR roles."
# The skill loads outreach-brief.md automatically and uses the context.

# Step 3: For complex decisions, invoke the outreach-strategist agent
# Type /agents and select outreach-strategist
# It reads outreach-brief.md and acts as Head of Growth, deciding which skills to run and in what sequence.

Key skill prompts

# campaign-angle-finder
"Run campaign-angle-finder for [CLIENT NAME].
Read outreach-brief.md for ICP and value props.
Generate three distinct angles, each leading with a different pain, proof point, or trigger.
For each angle: angle name, best-fit prospect type, recommended hook, and test batch recommendation.
Rule: run Angle 1 on 50 contacts first. Above 5% reply rate, scale it. Below 3%, move to Angle 2."

# outbound-campaign-architect
"Run outbound-campaign-architect for [CAMPAIGN NAME].
ICP from outreach-brief.md. Selected angle: [ANGLE NAME].
Design complete 6-touch sequence: channel and timing for each touch.
Target metrics: reply rate 6-8%, positive reply 2-3%, meetings per 100 contacts 2-3.
Include: size recommendations, exclusion list rules, and follow-up protocol for contacts who complete without replying."

# gtm-action-thinker (run before any new approach)
"Run gtm-action-thinker on this approach: [DESCRIBE APPROACH].
Four questions:
1. What assumption does this depend on and is it testable?
2. What happens if the assumption is wrong?
3. What is the cheapest way to validate before scaling?
4. What are the second and third order effects if this works?"

# copywriting-first-touch
"Run copywriting-first-touch. Prospect: [NAME], [TITLE] at [COMPANY].
Signal: [DESCRIBE]. Read outreach-brief.md for value props and ICP.
Rules: signal-led opening line, under 100 words, one question as CTA. No em dashes."

# reply-handler
"Run reply-handler on this reply: [PASTE REPLY].
Classify the reply type and return the full response script for this type."

# outbound-analyst
"Run outbound-analyst on this campaign data: [PASTE METRICS].
Benchmark against the 244,000+ campaign dataset.
Identify the stall stage and return a prioritised fix list."

Source: henriquecaner/next-level-outreach (7 skills for daily GTM content operations)

The content agent runs the daily GTM content operation: planning the day, pulling signals to Slack, extracting competitor intel from video, turning raw metrics into client reports, and writing articles that get cited by AI search. 7 skills. Most run in under 5 minutes.

The 7 skills and when to invoke each

/gtm-day-planner

Invoke first thing every morning. The skill asks five questions: active campaigns and statuses, what happened overnight, what you are most worried about, and any external deadlines before 12pm. Output: Tier 1 (fix before anything else, max 2 items), Tier 2 (move forward now, max 3 items), Tier 3 (defer to afternoon), Claude Code sessions to queue in Agent View, and the single most important thing to answer before doing anything else. Takes 90 seconds after the first run.

/gtm-digest

Invoke or schedule at 8am daily. Pulls campaign performance from Gmail alerts, client communications from Gmail and Slack, and ICP buying signals from Clay or n8n webhook entries. Posts a structured digest to a designated Slack channel. Maximum 25 lines. Only flags something urgent if it requires action before noon. Set up via Cowork scheduled task using Gmail and Slack connectors.

/watch

Invoke with a YouTube URL or any video transcript. The skill pulls the transcript and runs four analysis passes: tactic extraction (every specific tactic with implement/test/not applicable verdict), content hooks (3 shareable claims for LinkedIn), competitive intelligence (tool stack, client attraction angle, gap to own), and prospect intelligence (who watches this, what problem they are solving, the cold email first line it unlocks). Takes under 3 minutes per video.

/client-report

Invoke with raw campaign metrics or a call transcript. The skill runs a 5-question intake (client name, week's focus, wins to lead with, anything to exclude), extracts performance signals and narrative signals, applies a client language filter that translates GTM jargon into plain English, and writes the report in under 250 words. Includes a gut check that asks: would a non-GTM client understand every sentence.

/gtm-article

Invoke with any source content: a video via /watch, a LinkedIn post, a case study, a workflow walkthrough. The skill identifies the core question the article answers, writes an Answer Block in the first 100 words (the most important structural element for AI citation), structures the article with question-format H2s, adds evidence density (specific numbers and named examples throughout), and outputs a dated markdown file with frontmatter. Length targets: workflow walkthroughs 1,200-1,800 words, case studies 1,000-1,500 words.

/skill-creator

Invoke when you want to turn any GTM process into a repeatable skill. The skill runs a structured interview (5 base questions plus 2-3 follow-ups), then writes a complete SKILL.md with name, description, context, trigger, process, output, always, and never sections. After writing, it runs the skill on a test case and asks if anything needs adjusting before saving.

/decide

Invoke for any stack, hiring, or channel decision where being wrong costs more than two weeks of time or more than £1k. The skill runs a 7-question interview, then three agent passes: structurer (options with explicit assumptions and pros/cons), contrarian (holes, hidden assumptions, motivated reasoning), synthesizer (recommendation with reversibility and kill criteria). Output saves as a dated decision memo.

Setting up the content agent

Install: drop each skill file into .claude/skills/ or install via Cowork Settings > Customize > Skills > Upload. Each skill is a folder containing a SKILL.md file.

Schedule gtm-digest: in Cowork, create a New Task, add the digest instructions (replace CLIENT_EMAIL and CLIENT_CHANNEL_NAME with your values), set to daily at 8am, connect Gmail and Slack connectors, click Run Now to test before activating.

Add context to CLAUDE.md: the more your CLAUDE.md knows about your ICP, active clients, content calendar setup, and post format standards, the better every skill performs without re-briefing. At minimum: active client list, LinkedIn CTA format, banned words, and posting schedule.

For the /watch skill: run it against 3 competitor videos per week and 3 ICP-relevant thought leader videos. Feed the output into /gtm-article to turn competitor intelligence into published content in one session.

# Invoke each skill
/nl-outreach:gtm-day-planner
/nl-outreach:gtm-digest
/nl-outreach:watch [YOUTUBE URL]
/nl-outreach:client-report
/nl-outreach:gtm-article
/nl-outreach:skill-creator
/nl-outreach:decide

Source: henriquecaner/next-level-outreach (campaign skills) + gtmagents/gtm-agents (workflow orchestrators)

The campaign agent takes a brief and orchestrates the full campaign motion across multiple skills. Workflow orchestrators chain agents together so one command triggers a sequence of specialised agents working in parallel.

Three campaign skills to run before writing anything

Angle first. Architecture second. Copy third. This is the rule. Skipping angle generation and going straight to copy is the most common reason campaigns underperform.

campaign-angle-finder

Reads ICP and value props from outreach-brief.md and generates three distinct campaign angles, each leading with a different pain, proof point, or trigger. Full output includes: angle name, best-fit prospect type, recommended hook, and test batch recommendation. Rule: run Angle 1 on 50 contacts first. Above 5% reply rate, scale it. Below 3%, move to Angle 2 on a fresh 50. Never blend angles within the same campaign batch.

outbound-campaign-architect

Designs a complete campaign based on ICP, selected angle, send capacity, and previous performance data. Output covers: complete 6-touch sequence structure with channel and timing for each touch, target metrics (reply rate 6-8%, positive reply 2-3%, meetings per 100 contacts 2-3), size recommendations, exclusion list rules, and follow-up protocol for contacts who complete the sequence without replying.

gtm-action-thinker

Stress-tests any GTM idea or campaign approach before execution. Runs it through four questions: what assumption does this depend on and is it testable, what happens if the assumption is wrong, what is the cheapest way to validate before scaling, and what are the second and third order effects if this works. Run this before committing budget or list to any new approach.

Workflow orchestrators: chaining agents for full campaign execution

Orchestrators chain multiple specialised agents together. Instead of one agent doing one task, an orchestrator coordinates a team working in sequence. The campaign-orchestration orchestrator, for example, chains a content-strategist, designer, copywriter, email-marketer, social-media-manager, paid-media-specialist, analytics-engineer, and campaign-manager. One command, eight agents, full campaign from brief to distribution.

Four campaign orchestrators worth installing:

campaign-orchestration: full campaign launch, coordinates 8+ agents from content strategy through analytics.

content-pipeline-orchestration: ideation to distribution with editorial calendar, drafts, and channel delivery.

email-sequence-orchestration: full lifecycle email sequencing from trigger logic to copy to performance analysis.

intent-signal-orchestration: signal ingestion, prioritisation, and activation. Takes intent data and turns it into outreach sequences automatically.

# Install orchestrators
/plugin marketplace add gtmagents/gtm-agents
/plugin install campaign-orchestration
/plugin install abm-orchestration
/plugin install content-marketing
/plugin install email-marketing
/plugin install marketing-automation

# Run an orchestrator
"Run a full campaign for VP Sales at Series A SaaS targeting companies with active SDR job postings.
Use the signal angle. Target 50 contacts in the first batch."
# The orchestrator reads outreach-brief.md, selects the relevant skills, and runs them in sequence.

# ABM orchestration (8 coordinated agents)
/abm-orchestration:target-accounts --tier 1 --vertical "financial services"
# Builds: personalised landing page, outreach sequence, LinkedIn ads brief, analytics setup

# Intent signal activation
"Run intent-signal-orchestration. Signal source: [G2 BUYER INTENT / BOMBORA].
For each in-market account: classify tier, build connection note, draft email opener, route to sequence."

Source: henriquecaner/next-level-outreach (daily ops skills) + Cowork scheduled tasks

The daily ops agent runs recurring GTM tasks on a schedule without manual triggering. Morning digest, pipeline check, campaign performance review, inbox triage. Set up once. Runs automatically.

Four recurring ops to schedule

Morning GTM digest (7am daily)

Claude connects to Gmail and Slack, pulls campaign performance alerts and overnight data, surfaces buying signals from Clay or n8n, and posts a structured digest to your Slack channel. Maximum 25 lines. You open your laptop to a full brief rather than opening five tools. Setup: Cowork scheduled task + Gmail connector + Slack connector. Replace CLIENT_EMAIL and Slack channel name in the skill instructions before activating.

Weekly campaign performance review (Monday 8am)

Claude pulls raw metrics from your sequencer (Instantly, Smartlead, or similar via Gmail API alerts), benchmarks them against the 244,000+ campaign dataset in outbound-analyst.md, identifies the stall stage, and delivers a prioritised fix list. Output: what is broken and how to fix it in priority order. Setup: Cowork scheduled task, same Gmail connector.

Pipeline triage (Monday 8am)

Claude reads your pipeline data (CSV export from CRM), produces stage-by-stage conversion rates versus benchmarks, identifies which stage has the highest drop-off, flags deal stalls by average time in each stage, and outputs a recommended action for the stall stage. Can be chained with the campaign architect to auto-generate the next sequence for stalled contacts.

Inbox triage (daily at start of day)

Claude reads all unread emails from the last 24 hours, identifies the three most important based on sender domain (reads clients.txt from the project repo to know which domains are priority), drafts a reply for each, and posts a briefing to Slack with sender, one-sentence summary, and the subject line of each draft reply. Setup: Cowork scheduled task + Gmail connector + Slack connector + clients.txt file in the project repo.

How to set up scheduled agents in Cowork

Inside Cowork, click New Task. Name the task and add a one-line description. Write the full instructions. The more specific, the better the output. Set the schedule (daily, weekly, specific time). Click Run Now to test before activating. If the output looks wrong, update the instructions and run again before setting the schedule.

Make sure every connector the task needs is authorised before scheduling: Settings > Customize > Connectors > Add. Gmail and Slack cover most daily ops tasks. Add HubSpot or your CRM connector for pipeline tasks.

For tasks that reference a file (clients.txt for inbox triage, outreach-brief.md for campaign reviews), make sure the file exists in the project repo and the instructions reference the correct path.

Minimum schedule interval is 1 hour. For anything that needs to run more frequently, use the Claude Routines API trigger with an external cron job.

# Daily GTM digest (Cowork scheduled task at 7am)
"Connect to Gmail and Slack.
Pull: campaign performance alerts from last 24 hours, unread client messages, Clay webhook data.
Post a digest to #gtm-digest with:
- Three most urgent items (action required before noon)
- Campaign performance delta vs yesterday
- New buying signals detected
Maximum 25 lines. Flag nothing as urgent unless it genuinely requires action before noon."

# Weekly pipeline triage (Cowork scheduled task Monday 8am)
"Read pipeline-data.csv from the project repo.
Produce stage-by-stage conversion rates vs benchmarks from outbound-analyst.md.
Identify which stage has the highest drop-off rate.
For the stall stage: flag the specific deals stalling there and generate a recommended next action per deal.
Post to #pipeline-review."

# Inbox triage (Cowork scheduled task daily at 8am)
"Read all unread emails from the last 24 hours via Gmail connector.
Load clients.txt from the project repo to identify priority sender domains.
For the three most important emails: draft a reply, rate the sender's intent (hot/warm/cold), and flag if action is needed today.
Post to #inbox-triage with: sender, one-sentence summary, reply draft filename, and action flag."

Source: Claude Agents for B2B (5 agents for complex multi-stakeholder deals)

Five Claude agents that handle the B2B deal journey: lead qualification, account research, client onboarding, competitive intelligence, and proposal generation.

B2B agents focus on the deal itself. GTM agents focus on pipeline generation. RevOps agents focus on data and reporting. The difference: B2B deals involve 5-10 stakeholders, 3-6 month sales cycles, technical evaluation, procurement, and legal review. Agents handle this context volume better than humans.

The lead qualification agent

The lead qualification agent connects Claude to your lead data, scores each lead against your ICP, and routes high-priority leads to Slack and premium sequences automatically. No manual qualification step.

How it works: three stages.

Stage 1: trigger and enrich. The workflow triggers from your CRM on a schedule or via webhook. Claude connects to Apollo or ZoomInfo to enrich each lead with firmographic and technographic data. The enriched data is structured before passing to the AI node.

Stage 2: score. The AI Agent node receives the enriched lead data along with a system prompt containing your ICP qualification criteria. The prompt uses dynamic injection: your ICP definition, trigger signals, and scoring thresholds are injected into each prompt automatically from your brief file. Claude (or your chosen model) returns a JSON object with an ICP score and a one-sentence reasoning.

Stage 3: route. An IF node reads the JSON score. Score 70 or above: route to Slack alert channel and enroll in premium sequence via Instantly. Below 70: route to nurture sequence. All leads: update CRM record with score and log to Google Sheets. The entire workflow runs without a single manual step.

Building it with the n8n-workflow-builder skill:

"Build a lead qualification workflow that triggers from a CRM webhook,
enriches each lead via Apollo,
sends enriched data to Claude with an ICP qualification prompt,
parses the JSON score,
routes leads above 70 to a Slack alert and Instantly premium sequence,
routes the rest to a nurture sequence,
and logs everything to Google Sheets."
# The n8n-workflow-builder skill outputs complete JSON you import directly into n8n.
# Add your API keys to the credential nodes, test with one lead, then activate.

Model selection for scoring:

• Claude Sonnet: approximately $0.15 per 1,000 leads. Best for complex qualification logic with nuanced criteria.
• GPT-4 Mini: approximately $0.01 per 1,000 leads. Good for speed-sensitive high-volume scoring.
• Gemini Flash: approximately $0.01 per 1,000 leads. Strong for scoring that involves market mapping or exploratory criteria.

For high-volume qualification (10,000+ leads per month), use a cheaper model for the first-pass score and Claude only for leads that pass the first threshold.

The account research agent

The account research agent produces deep intelligence on any target account before a sales interaction. Takes under 5 minutes. Manual equivalent takes 45-60 minutes and is rarely done as thoroughly.

Six sections in every brief:

Company overview. Current positioning, what they sell and to whom, recent news, funding stage, headcount and growth trajectory, and notable client names if publicly visible. Relevant because the opening line of any sales interaction should reference something specific about the company, not a generic observation.

Stakeholder map. Key decision-makers and influencers at the account: their roles, tenure, stated priorities from LinkedIn activity and public content, and their likely relationship to the problem you solve. The deep-company-analyser skill surfaces the internal language the company uses (from job postings and published content) so your language matches theirs.

Tech stack. Tools currently in use (via BuiltWith or similar signal sources), recently added tools, and recently dropped tools. Tech stack changes are buying signals: a CRM migration, a new data stack, a recent SDR tooling hire all indicate active evaluation cycles.

Buying signals. Recent trigger events: job postings that indicate the problem you solve, funding announcements, leadership changes, expansion into new markets, and product launches that create new adjacent pain. The trigger-finder skill maps each signal type to a detection method and an urgency window.

Angle recommendations. Based on the signals found, two or three specific outreach angles ranked by strength. Not generic pain points: specific observations tied to specific signals. The deep-company-analyser skill produces these automatically when it reads the account data.

Prep questions. Three questions to ask in the first call, specific enough to demonstrate genuine research and open the conversation toward the strongest angle.

# Run the account research agent
"Run deep-company-analyser on [COMPANY NAME or DOMAIN].
Focus: their outbound and SDR motions. I sell [DESCRIBE OFFER IN ONE SENTENCE].
Return the full six-section brief."

# For volume account research (20+ accounts before a campaign)
# Spawn 4-5 parallel Claude Code sessions in Agent View, each handling a batch.
# Total time for 20 accounts is roughly the time it takes to do one manually.

# For named accounts: run this agent the evening before any call.
# For Dispatch: send the task from your phone in the evening, specify company name and call time.
# The brief is on your desktop when you arrive.

The client onboarding agent

The client onboarding agent automates the full onboarding workflow: folder creation, brief ingestion, Linear project setup, Slack notification, and brief-populated template generation. What used to take 45 minutes now takes under 2 minutes.

Five automated steps from signed contract to ready-to-start project:

Step 1: create the client directory. The agent creates a folder at clients/[client-name]/ with the standard structure: outreach-brief.md (populated from new-client.json intake data), memory.md, campaign/ folder, reports/ folder, and assets/ folder.

Step 2: populate the brief. The agent reads new-client.json (the intake form completed during onboarding) and fills all placeholder sections of the outreach-brief.md template: company description, ICP trigger definition, personas, value props, competitors, and buying signals. The filled brief is immediately usable by the outreach agent.

Step 3: create the Linear project. The agent creates a Linear project in the GTM team with the client name, adds the standard task set (outreach-brief.md review, first campaign setup, first report), and assigns according to your team's routing rules.

Step 4: send Slack notification. Posts to #new-clients with: client name, brief location, Linear project link, and the three most important things from the brief that the team should know before starting.

Step 5: rollback handling. If Linear or Slack steps fail, the agent does not silently proceed. It reports the failure and waits. The client directory and brief are preserved regardless of whether downstream steps complete. Build rollback into the plan explicitly before coding this agent.

# Build using Explore-Plan-Code-Commit pattern
# MCP connections needed: Linear MCP + Slack MCP (both in Cowork connectors directory)

# Step 1: explore existing onboarding scripts and client data structure
# Step 2: plan with ultrathink (automation creates directory, copies and populates template, creates Linear project, sends Slack notification, with rollback if any step fails)
# Step 3: code steps 1-2 first, test on a sample new-client.json fixture, then implement Linear and Slack steps
# Step 4: sub-agent verifies rollback exists for Linear and Slack steps and that no existing client directory would be overwritten

# new-client.json intake form fields:
# company_name, icp_description, known_competitors, primary_value_prop, key_contacts

The competitive intelligence agent

Four skills from the GTM Agents marketplace. Run continuously for ongoing competitive monitoring or on-demand for specific deal preparation.

Competitive landscape mapping. Produces a structured map of the competitive landscape: direct competitors, adjacent tools, and emerging entrants. For each competitor: positioning, target customer, pricing model, key differentiators, and known weaknesses. Output format is a comparison table your sales team can reference before any competitive deal.

Battlecard development. Takes the landscape map and produces a one-page battlecard per competitor: their positioning versus yours, the three questions to ask that reveal their weaknesses, the three objection responses when a prospect mentions them by name, and the proof point that wins the head-to-head comparison. Battlecards update automatically when you re-run the landscape map and detect changes.

Win/loss interview framework. Generates a structured interview guide for win/loss conversations: opening questions to establish rapport and context, questions that surface the real decision criteria (not the stated ones), questions that reveal where the competitor won or lost on specific dimensions, and closing questions that identify what would have changed the outcome. The framework adjusts based on whether it is a win or loss interview.

Competitor monitoring. A scheduled agent that checks competitor websites, job postings, LinkedIn updates, and G2 reviews on a weekly cadence. Flags: new features announced, pricing changes, leadership changes, new customer segments being targeted, and negative patterns in review sentiment. Delivers a digest to Slack every Monday morning alongside the pipeline review.

# Install the competitive intelligence skills
/plugin install sales-enablement

# Landscape mapping
/sales-enablement:battlecard --competitor "[COMPETITOR NAME]"
# Returns: their positioning vs yours, 3 questions that reveal their weaknesses,
# 3 objection responses, the proof point that wins the head-to-head

# Scheduled competitor monitoring (Cowork weekly task, Monday 7am)
"Check competitor websites, job postings, LinkedIn updates, and G2 reviews for [COMPETITOR LIST].
Flag: new features, pricing changes, leadership changes, new customer segments, negative review patterns.
Post digest to #competitive-intel."

# Win/loss interview framework
"Generate a win/loss interview guide for a deal we [WON / LOST] against [COMPETITOR].
Focus: surface the real decision criteria, not the stated ones.
Include closing questions that identify what would have changed the outcome."

The proposal generation agent

Reads the account brief and generates a structured proposal in under 5 minutes. Built using the Explore-Plan-Code-Commit pattern.

The proposal generator reads clients/[client-name]/outreach-brief.md and any call notes in the client folder and generates a proposal at proposals/[client-name]-[date].md.

Proposal structure:

• Situation summary: what you learned about the account and their specific problem
• Proposed approach: the outbound or GTM motion matched to their ICP and buying signals
• Timeline: three phases with specific deliverables per phase
• Pricing: tier options with clear scope definitions for each

Output is clean markdown suitable for PDF conversion.

# Run the proposal generator
"Read clients/[CLIENT NAME]/outreach-brief.md and all call notes in the client folder.
Generate a proposal at proposals/[CLIENT NAME]-[DATE].md.

Structure:
- Situation summary: what we learned about their specific problem
- Proposed approach: the GTM motion matched to their ICP and buying signals from the brief
- Timeline: three phases with specific deliverables per phase
- Pricing: two tier options with clear scope definitions for each

Output: clean markdown. Convert to PDF with the pdf skill or Pandoc."

# Convert to PDF
"Convert proposals/[CLIENT NAME]-[DATE].md to a client-ready PDF.
Use clean formatting: cover page with client name and date, section headers, clean body text."

# Always test against a sample outreach-brief.md fixture before running on a real client.
# Output quality depends directly on brief quality.
# If the brief has vague ICP or missing value props, the proposal will reflect that.

Sources: github.com/gtmagents/gtm-agents · github.com/alirezarezvani/claude-skills · github.com/VoltAgent/awesome-agent-skills · github.com/coreyhaines31/marketingskills · github.com/sales-skills/sales · github.com/zubair-trabzada/ai-sales-team-claude · github.com/louisblythe/Sales-Skills · github.com/ComposioHQ/awesome-claude-skills · github.com/indranilbanerjee/digital-marketing-pro · github.com/modest-curator478/claude-skills · github.com/BehiSecc/awesome-claude-skills

All agents in this section work with Claude Code, Cursor, Codex CLI, Gemini CLI, and any harness that reads SKILL.md files.

Sales and prospecting agents

ai-sales-team-claude (14 skills, 5 parallel agents)

github.com/zubair-trabzada/ai-sales-team-claude

Research prospects, qualify leads (BANT + MEDDIC), find decision makers, generate outreach sequences, prepare for meetings, write proposals, produce PDF pipeline reports.

curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/ai-sales-team-claude/main/install.sh | bash
/sales prospect https://stripe.com       # 5 agents fire simultaneously, returns fit score, decision makers, BANT, competitors, sequence
/sales qualify https://notion.so         # BANT + MEDDIC scoring
/sales outreach "Linear"                 # 5-email signal-led sequence
/sales prep https://datadog.com          # pre-call brief
/sales proposal "Client Name"           # full proposal
/sales competitors https://client.com   # competitive positioning
/sales icp "B2B SaaS, 50-200 employees" # ICP definition

gtmagents sales plugins (67 plugins, 92 agents, 263 skills)

github.com/gtmagents/gtm-agents

/plugin marketplace add gtmagents/gtm-agents
/plugin install sales-prospecting     # 5 skills: list building, ICP, contact discovery, account tiering, sequencing
/plugin install sales-pipeline        # pipeline health, forecast, deal inspection, coverage
/plugin install enterprise-sales      # stakeholder map, champion plan, value stories, procurement templates
/plugin install sales-calls           # pre-call brief, live call prompts, post-call analysis, follow-up
/plugin install sales-coaching        # rep outreach audits, coaching plans, cohort reviews
/plugin install account-management    # success plans, QBRs, expansion signals
/plugin install sales-enablement      # battlecards, competitive intel, playbooks, launch plans
/plugin install sales-operations      # territory planning, capacity planning, comp plan design

louisblythe/Sales-Skills (B2B sales for SDRs and AEs)

github.com/louisblythe/Sales-Skills

npx add-skill louisblythe/salesskills
# Skills: discovery-call, pricing-negotiation, competitor-alternatives, decision-maker-mapping, pipeline-review, sales-playbook

buyer-eval (structured B2B vendor evaluation)

github.com/salespeak-ai/buyer-eval-skill

Researches your company, asks domain-expert questions, scores vendors across 7 dimensions, produces comparative scorecards with demo prep questions.

git clone https://github.com/salespeak-ai/buyer-eval-skill ~/.claude/skills/buyer-eval-skill
"Using buyer-eval, evaluate [VENDOR] as a [CATEGORY] solution. Score across 7 dimensions. Produce comparative scorecard."

hubspot-admin-skills (audit, clean, enrich, automate HubSpot)

npx skills add hubspot-admin-skills
"Run hubspot-admin-skills. Audit my HubSpot CRM. Find: contacts without companies, deals with no next step, stale records, duplicate companies, stage mismatches. Return a prioritised fix list and weekly maintenance checklist."

Composio CRM automation (Close, HubSpot, Pipedrive, Salesforce, Zoho)

github.com/ComposioHQ/awesome-claude-skills

/plugin marketplace add ComposioHQ/awesome-claude-skills
# Automate any CRM via Rube MCP: leads, contacts, opportunities, activities, pipelines
"Using the HubSpot automation skill: find all deals in Discovery stage older than 14 days. Create a follow-up task for each owner. Update probability to 60%. Log the action."
"Using the Salesforce automation skill, move all Opportunity records where CloseDate is in the past and Stage is not Closed. Set Stage = At Risk. Create a follow-up task for the owner."

Outbound and cold email agents

coreyhaines31/cold-email (signal-led sequences)

github.com/coreyhaines31/marketingskills

npx skills add coreyhaines31/marketingskills --skill cold-email
"Write a cold email sequence for VP Sales at B2B SaaS companies (50-500 employees). Signal: they just posted a Head of SDR role. Offer: [ONE SENTENCE]. 5 touches, different angle per touch, subject line variants per email. No em dashes."

gtmagents/cold-email-personalization (6-step workflow, QA gate at 80+)

/plugin install cold-email-personalization
/cold-email-personalization:signal-research "Acme Corp" --contact "Sarah Chen, VP Sales"
/cold-email-personalization:qa-score --email email-draft.txt
# Gate: score must hit 80+ before sending

sales-skills/sales (GTM suite: CRM, outbound, enrichment, social listening)

github.com/sales-skills/sales

npx skills add sales-skills/sales
"Using the social-listening skill, monitor [TOPIC / COMPETITOR / PAIN POINT] across Reddit, LinkedIn, X, GitHub, HN. For each result: platform, URL, date, intent score, recommended action."

Marketing and growth agents

coreyhaines31/marketingskills (40+ SaaS marketing skills)

github.com/coreyhaines31/marketingskills

npx skills add coreyhaines31/marketingskills
# Individual installs via VoltAgent:
npx skills add coreyhaines31/ab-test-setup
npx skills add coreyhaines31/ad-creative
npx skills add coreyhaines31/ai-seo
npx skills add coreyhaines31/analytics-tracking
npx skills add coreyhaines31/churn-prevention
npx skills add coreyhaines31/competitor-alternatives
npx skills add coreyhaines31/content-strategy
npx skills add coreyhaines31/copywriting
npx skills add coreyhaines31/email-sequence
npx skills add coreyhaines31/growth-loops
npx skills add coreyhaines31/launch-strategy
npx skills add coreyhaines31/lead-magnet
npx skills add coreyhaines31/onboarding-activation
npx skills add coreyhaines31/paywall-optimisation
npx skills add coreyhaines31/pr-strategy
npx skills add coreyhaines31/prospect-list-building
npx skills add coreyhaines31/referral-program
npx skills add coreyhaines31/revops
npx skills add coreyhaines31/sales-collateral
npx skills add coreyhaines31/seo-audit

digital-marketing-pro (158 skills, 25 agents, 12-Part Strategy Flow)

github.com/indranilbanerjee/digital-marketing-pro

Cowork team-persistent. For agencies running 50+ brands. AEO/GEO across 6 platforms. EU AI Act Article 50 ready.

npx skills add indranilbanerjee/digital-marketing-pro
/digital-marketing-pro:cowork-setup  # team setup: brand state persists via Google Drive
/dmp:brand-discovery
/dmp:icp-research
/dmp:competitor-map
/dmp:content-strategy
/dmp:seo-aeo-audit    # checks ChatGPT, Perplexity, Gemini, Copilot, AI Mode, Bing
/dmp:paid-strategy
/dmp:email-strategy
/dmp:cro-audit
/dmp:analytics-setup
/dmp:execution-calendar  # 90-day plan

alirezarezvani/marketing-skills (43 skills, 8 pods)

github.com/alirezarezvani/claude-skills

/plugin install marketing-skills@claude-code-skills
# Pods: Content, SEO + AEO, CRO, Channels, Growth, Intelligence, Sales
# AEO: E-E-A-T audit, citation tracking across 5 LLMs
# Brand voice analysis:
python3 marketing-skill/content-production/scripts/brand_voice_analyzer.py article.txt

gtmagents marketing plugins

/plugin install content-marketing      # 6 writing skills, editorial calendar, distribution
/plugin install email-marketing        # lifecycle sequences, drip, re-engagement
/plugin install marketing-automation   # lead scoring, nurture, workflow triggers
/plugin install abm-orchestration     # 8-agent ABM campaign builder

Product and SaaS agents

alirezarezvani/product-skills (12 skills including senior PM agent)

github.com/alirezarezvani/claude-skills

Senior PM agent with 6 knowledge domains, 12 templates, 30+ frameworks. Discovery, strategy, delivery, SaaS metrics, PM career coaching, AI product craft.

/plugin install product-skills@claude-code-skills
# RICE prioritisation:
python3 product-team/product-manager-toolkit/scripts/rice_prioritizer.py features.csv
# SaaS scaffolder, experiment designer, analytics skill, landing page generator, roadmap communicator

BehiSecc/product-manager-skills

github.com/BehiSecc/awesome-claude-skills

Senior PM agent with 6 knowledge domains, 12 templates, and 30+ frameworks.

npx skills add product-manager-skills
"Using product-manager-skills, run a discovery session for [PROBLEM AREA]. Framework: Torres Opportunity Solution Tree."

coreyhaines31 product skills

npx skills add coreyhaines31/onboarding-activation   # post-signup onboarding, time-to-value
npx skills add coreyhaines31/paywall-optimisation     # in-app paywalls, upsell modals
npx skills add coreyhaines31/free-tool-strategy       # lead gen via free tools
npx skills add coreyhaines31/churn-prevention         # cancellation flows, save offers, failed payment recovery

CRM, RevOps, and finance agents

modest-curator478/claude-skills (HubSpot, ABM, pipeline, win/loss)

github.com/modest-curator478/claude-skills

git clone https://github.com/modest-curator478/claude-skills
cp -R claude-skills/skills/* ~/.claude/skills/
"Using the win-loss skill, analyse our last 20 closed opportunities. Find: win rate by ICP segment, loss reasons by frequency, the one change most likely to improve win rate."

elijeangilles/revops-skills (Salesforce pipeline audit, A-F grading)

github.com/elijeangilles/revops-skills

git clone https://github.com/elijeangilles/revops-skills
cp -R revops-skills/skills/* ~/.claude/skills/
"Run salesforce-revops-audit. Audit across 5 dimensions: pipeline health, data quality, forecast hygiene, deal integrity, process integrity. Return overall grade A-F, top 5 flagged deals, remediation skills to run next."

alirezarezvani/finance-skills (SaaS metrics coach)

/plugin install finance-skills@claude-code-skills
python3 finance/saas-metrics-coach/scripts/metrics_calculator.py --mrr 80000 --customers 200 --churned 3 --json
# Returns: all 7 core SaaS metrics vs benchmarks, one lever to prioritise this quarter

alirezarezvani/c-level-skills (28 C-suite advisors)

/plugin install c-level-skills@claude-code-skills
# CRO advisory, CMO GTM strategy, CFO fundraise prep, CTO tech debt scoring
python3 c-level-advisor/cto-advisor/scripts/tech_debt_analyzer.py /path/to/codebase

Customer success and retention agents

gtmagents/account-management (success plans, QBRs, expansion)

/plugin install account-management
/account-management:qbr-prep "Enterprise Customer" --contract 200000 --renewal 2026-12-01
/account-management:expansion-signals --account "Acme Corp"
/account-management:churn-risk-review --segment enterprise --period this-month

coreyhaines31/churn-prevention

npx skills add coreyhaines31/churn-prevention
"Build a churn prevention system for [PRODUCT]. Health score: 5 product signals + 3 non-product signals. Save sequence: in-app, email, CS call. Failed payment recovery: 5-touch sequence."

alirezarezvani/cs-pm-orchestrator (continuous discovery, Torres OST)

/cs:pm  # routes between discovery, roadmap, success plans, experiment design, metrics review
"Lint this Opportunity Solution Tree: [PASTE]. Check: is each opportunity framed as a customer outcome? Are experiments mapped to assumptions? Are solutions attached to opportunities?"

Multi-agent frameworks and orchestration

gtmagents full suite (67 plugins, 263 skills, 92 agents)

github.com/gtmagents/gtm-agents

/plugin marketplace add gtmagents/gtm-agents
# Full orchestration:
"I am a [ROLE] at a [STAGE] B2B SaaS company selling [PRODUCT] to [ICP].
Current ARR: [$N]. Goal: [$N] by [DATE].
Orchestrate: what to do in sales, marketing, and CS this quarter to hit the goal."

alirezarezvani full suite (345 skills, 30+ agents, 17 domains)

github.com/alirezarezvani/claude-skills

/plugin marketplace add alirezarezvani/claude-skills
/plugin install product-skills@claude-code-skills
/plugin install marketing-skills@claude-code-skills
/plugin install c-level-skills@claude-code-skills
/plugin install finance-skills@claude-code-skills
/plugin install pm-skills@claude-code-skills

VoltAgent/awesome-agent-skills (1,000+ community skills)

github.com/VoltAgent/awesome-agent-skills

Compatible with Claude Code, Codex, Antigravity, Gemini CLI, Cursor, GitHub Copilot, OpenCode, Windsurf.

# Install any coreyhaines31, google-labs, googleworkspace, duckdb skills via VoltAgent:
npx skills add [org]/[skill-name]

manage-skills (cross-agent skill manager: 11 tools)

github.com/rohitg00/awesome-claude-code-toolkit

Discover, create, edit, toggle, copy, move, and delete agent skills across Claude Code, Cursor, Windsurf, Copilot, Codex, Cline, Aider, Continue, Roo Code, Augment.

cp -r skills/manage-skills ~/.claude/skills/manage-skills
"Run manage-skills. List all my installed skills across Claude Code, Cursor, and Codex."
"Using manage-skills, audit all installed skills. Flag any that are outdated or conflicting."

Every install command and critical prompt from this playbook in one place. No em dashes anywhere in any output.

Core install sequence

# GTM agents (outreach, content, campaign, daily ops)
/plugin marketplace add henriquecaner/next-level-outreach
/plugin install nl-outreach@nl-outreach-marketplace

# B2B deal agents (via gtmagents and direct install)
/plugin marketplace add gtmagents/gtm-agents
/plugin install sales-calls
/plugin install account-management
/plugin install sales-enablement

# Lead qualification workflow
# Build via n8n-workflow-builder skill + import JSON into n8n

# Client onboarding agent (build using Explore-Plan-Code-Commit)
# MCP: Linear MCP + Slack MCP (both in Cowork connectors directory)

# 50+ agent directory
curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/ai-sales-team-claude/main/install.sh | bash
npx add-skill louisblythe/salesskills
npx skills add coreyhaines31/marketingskills
npx skills add sales-skills/sales
npx skills add indranilbanerjee/digital-marketing-pro
/plugin marketplace add alirezarezvani/claude-skills
/plugin install product-skills@claude-code-skills
/plugin install marketing-skills@claude-code-skills
/plugin install c-level-skills@claude-code-skills
/plugin install finance-skills@claude-code-skills
/plugin marketplace add ComposioHQ/awesome-claude-skills
git clone https://github.com/elijeangilles/revops-skills && cp -R revops-skills/skills/* ~/.claude/skills/
git clone https://github.com/modest-curator478/claude-skills && cp -R claude-skills/skills/* ~/.claude/skills/

The outreach agent quick start

# 1. Onboard (once per client)
/nl-outreach:onboard

# 2. Find the campaign angle
/nl-outreach:campaign-angle-finder

# 3. Build the campaign structure
/nl-outreach:outbound-campaign-architect

# 4. Write the first touch
/nl-outreach:copywriting-first-touch

# 5. Write the LinkedIn sequence
/nl-outreach:linkedin-sequence

# 6. Stress-test before sending
/nl-outreach:gtm-action-thinker

# 7. Analyse replies
/nl-outreach:reply-handler

# 8. Review pipeline
/nl-outreach:pipeline-analysis

The B2B deal agent quick start

# Before any sales call: account research agent
"Run deep-company-analyser on [COMPANY]. Focus: [THEIR PAIN RELEVANT TO YOUR OFFER].
Return: company overview, stakeholder map, tech stack, buying signals, angle recommendations, prep questions."

# After winning a deal: client onboarding agent
# Reads new-client.json, creates folder structure, populates outreach-brief.md,
# creates Linear project, posts to #new-clients Slack channel

# Before a competitive deal: competitive intelligence agent
/sales-enablement:battlecard --competitor "[COMPETITOR]"

# After a discovery call: proposal agent
"Read clients/[CLIENT]/outreach-brief.md and call-notes.md.
Generate a proposal at proposals/[CLIENT]-[DATE].md."

# For lead qualification at scale: build n8n workflow
"Using n8n-workflow-builder: build a lead qualification workflow.
Trigger: CRM webhook. Enrich: Apollo. Score: Claude with ICP prompt. Route: 70+ to Slack + Instantly premium, below 70 to nurture. Log: Google Sheets."

The daily and weekly agent routines

# Daily morning (Cowork 7am scheduled task)
"Connect to Gmail and Slack. Pull: campaign alerts, client messages, buying signals from Clay.
Post digest to #gtm-digest. Max 25 lines. Flag nothing as urgent unless action needed before noon."

"Triage LinkedIn inbox. Classify: positive / neutral / negative.
For each positive: draft reply moving toward a call. Show drafts. Wait for approval."

# Weekly Monday (Cowork 8am scheduled task)
"Pull raw metrics from [SEQUENCER]. Benchmark against outbound-analyst.md 244k campaign dataset.
Identify the stall stage. Return prioritised fix list."

"Read pipeline-data.csv. Stage-by-stage conversion vs benchmarks. Flag stalls.
For each stalled deal: recommended next action. Post to #pipeline-review."

# Friday review
"Review this week: email open/reply/positive rates, LinkedIn engagement, outreach connections and conversations, pipeline deals moved.
Tell me the one thing to do differently next week."

Model routing for agents

# Orchestrators and complex reasoning: claude-sonnet-5 at medium effort
# Sub-agents doing data gathering: claude-haiku-4-5 (fast, cheap)
# Strategy, copy, competitive analysis: claude-opus-4-8
# Lead qualification at volume: GPT-4 Mini or Gemini Flash ($0.01 per 1,000 leads)

# Cost routing for qualification:
# First-pass score (simple 3-criteria ICP): Gemini Flash or GPT-4 Mini
# Leads that pass first threshold (complex scoring): Claude Sonnet

# Fable 5 for:
# Reading past transcripts to diagnose your setup and generate new agents
# Long-horizon autonomous execution with /goal
# Run session audit: "Reflect on my past Claude Code sessions.
# Cluster signals. For each cluster: new agent / new automation / fix / nothing.
# Write candidates to improvements.md. Diagnosis only first."