{"id":"424677e8-0dbb-44bf-89c0-b0aa6ef6bd1c","slug":"firsttouch-agent-skill-packs","kind":"skill","name":"firsttouch-agent-skill-packs","summary":"FirstTouch Agent Skill Packs for Claude, Cursor, Windsurf, ChatGPT, and MCP-compatible agents","owner":"First-Touch-Inc","repoUrl":"https://github.com/First-Touch-Inc/firsttouch-agent-skill-packs","license":"MIT","licenseTier":"permissive","trust":"community","stars":5,"match":null,"fewStars":false,"format":"skill","createdAt":"2026-06-29T17:10:10.000Z","addedAt":"2026-10-21T09:12:07.609Z","logoUrl":"https://avatars.githubusercontent.com/u/244586852?v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["writing"],"tags":["ai-sdr","claude","hubspot","linkedin","sales-automation"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-26T04:46:19.305Z","sample":false,"url":"https://conduits.dev/tools/firsttouch-agent-skill-packs","language":"Python","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# FirstTouch Revenue Skills - Customer Pack\n\n> **Turn Claude, Cursor, Codex, and ChatGPT into FirstTouch-native LinkedIn revenue operators.**\n\nThis pack gives customers a structured set of **plays** - reusable skills that execute real LinkedIn-outreach workflows with FirstTouch, human approvals, account-safety limits, and HubSpot context when available.\n\nIt is built to the open **Agent Skills** standard ([agentskills.io](https://agentskills.io)), so every skill works in Claude Code, Claude.ai, Cursor, Windsurf, and any MCP-compatible harness.\n\n---\n\n## Why this pack exists\n\nMost \"AI outreach\" tools stop at generating copy. Customers can get an agent to *write* a message, but not to:\n\n- act on a warm signal like post engagement, a form fill, a deal stall, or a website visit\n- find the right person on LinkedIn and reach out safely\n- prepare approval-ready FirstTouch queues instead of autonomous sends\n- log and route through HubSpot when CRM access is connected\n- respect account-safety limits and human approval gates\n\nEach play closes that loop. This is the difference between an agent that *sounds* like an SDR and one that *operates* like one.\n\n---\n\n## How the pack is organized\n\n```\nfirsttouch-skills/\n├── README.md                              ← you are here\n├── references/\n│   ├── onboarding.md                      ← first-run questions, play catalog, account limits\n│   ├── system-grounding.md                ← how FirstTouch + HubSpot + agents fit together\n│   ├── mcp-setup.md                       ← connect HubSpot MCP, FirstTouch MCP, enrichment MCPs\n│   ├── messaging-framework.md             ← FirstTouch LinkedIn messaging methodology\n│   ├── safety-governance.md               ← approval gates, send safety, account limits\n│   └── troubleshooting.md                 ← MCP failures, cooldowns, recovery\n└── skills/\n    ├── firsttouch-messaging/              ← foundation: how to write outreach\n    ├── warm-engager-followup/             ← social engagement flow\n    ├── social-campaigns/                  ← one-time static-template campaigns\n    ├── icp-outbound-builder/              ← AI SDR\n    ├── founder-led-outbound/              ← founder AI SDR\n    ├── inbound-speed-to-lead/\n    ├── hubspot-signal-to-linkedin-touch/\n    ├── website-visitor-followup/\n    ├── stalled-deal-reactivation/\n    ├── customer-referral/\n    ├── hubspot-social-task-runner/        ← execute due HubSpot social tasks\n    ├── team-performance-report/           ← FirstTouch team metrics + attribution\n    ├── sequence-qa-reviewer/\n    ├── workspace-audit/\n    ├── campaign-pause-and-fix/            ← mid-send escape hatch: pause, fix, restart\n    ├── founder-post-generator/            ← posts that create the engagement to harvest\n    ├── post-demo-followup/                ← same-day demo follow-up + stakeholder expansion\n    └── champion-job-change/               ← champion moved: new-company path + old-account protection\n```\n\n**Two layers:**\n1. **Foundation** (`references/` + `firsttouch-messaging`) - shared grounding every play assumes.\n2. **Customer plays** - self-contained, revenue-tied workflows. Governance and attribution rules live in shared references instead of separate customer-facing plays.\n\n---\n\n## Get started in 5 steps\n\n**Claude Code users: skip the steps below** - install your pack as a plugin and the FirstTouch MCP connects automatically:\n\n```\n/plugin marketplace add First-Touch-Inc/firsttouch-agent-skill-packs\n/plugin install founder-pack@firsttouch        # or ae-pack / bdr-pack / revops-pack\n```\n\nThen run `/mcp` once to authorize the FirstTouch server (a \"Needs authentication\" status before that is expected), answer the onboarding questions when the agent asks, and run your first play. Everyone else:\n\n1. **Download your pack** (below) - or grab individual skills from [`packs/skills/`](packs/skills/).\n2. **Install it** for your agent - platform guides in [`install/`](install/): [Claude Code](install/claude-code.md) · [Claude.ai](install/claude-ai.md) · [Cursor/Windsurf](install/cursor-windsurf.md) · [ChatGPT](install/chatgpt.md).\n3. **Connect the FirstTouch MCP** (`https://mcp.firsttouch.ai`) - required by every play. HubSpot MCP is optional but unlocks the CRM plays. See [`references/mcp-setup.md`](references/mcp-setup.md).\n4. **Answer the 3 onboarding questions** in [`references/onboarding.md`](references/onboarding.md) - LinkedIn account type, HubSpot access, which play to run first.\n5. **Run your first play and approve the drafts** - nothing sends without your approval.\n\n### Download your pack\n\n| Pack | For | Download |\n|---|---|---|\n| Founder Pack | Founders doing their own sales | [`packs/founder-pack.zip`](packs/founder-pack.zip) |\n| AE Pack | Account Executives | [`packs/ae-pack.zip`](packs/ae-pack.zip) |\n| BDR Pack | Business Development Reps | [`packs/bdr-pack.zip`](packs/bdr-pack.zip) |\n| RevOps Pack | Revenue Operations | [`packs/revops-pack.zip`](packs/revops-pack.zip) |\n\nSingle skills for Claude.ai (one skill per zip, references bundled): [`packs/skills/`](packs/skills/)\n\n---\n\n## Install (for end customers)\n\n1. Drop any skill folder into your agent's skills directory:\n   - **Claude Code:** `~/.claude/skills/<skill-name>/` (or `.claude/skills/` per-project)\n   - **Claude.ai:** Settings → Features → upload a single-skill zip from [`packs/skills/`](packs/skills/)\n   - **Cursor / Windsurf / custom:** copy the folder anywhere the agent can read\n2. Connect the MCPs your plays need - see [`references/mcp-setup.md`](references/mcp-setup.md).\n   - **FirstTouch MCP** is required for FirstTouch execution and approvals.\n   - **HubSpot MCP or a service key/private app token** unlocks HubSpot-specific plays.\n   - If HubSpot is unavailable, run FirstTouch-only plays such as social engagement, social campaigns from Discover Contacts or imported lists, AI SDR from Discover Contacts, founder AI SDR, sequence QA, and workspace audit.\n3. Complete [`references/onboarding.md`](references/onboarding.md) before running the first play.\n4. Start with high-intent plays first; add outbound only after those are running and account limits are respected.\n\n---\n\n## First-run onboarding and recommended rollout\n\nEvery pack should ask three questions before running volume:\n\n| Question | Why it matters |\n|---|---|\n| Do you have Sales Navigator / Premium, or a free/basic LinkedIn account? | Free/basic accounts cannot use connection notes; recommend **10** connection requests/day (FirstTouch max 20/day). Sales Navigator / Premium can use notes; recommend **20**/day (FirstTouch max 30/day). |\n| Do you use HubSpot, and can you connect the HubSpot MCP or provide a service key/private app token? | HubSpot-specific plays need internal CRM context. Without it, run FirstTouch-only plays or ask the user to create a HubSpot list/source FirstTouch can access. |\n| Which plays do you want to run first? | Recommend high-intent plays first, then outbound once warm motions are running to keep the account healthy. |\n\n**Recommended order:**\n1. For founders, start with **Social Engagement Flow** first. It does not require HubSpot and uses the warmest signal available.\n2. Run other high-intent plays next: inbound speed-to-lead, website visitor follow-up, HubSpot signal touches, customer referral thank-you, and stalled-deal reactivation when the required data exists.\n3. Add one-time social campaigns when the audience is narrow and the user can approve static flow templates.\n4. Add AI SDR / Founder AI SDR only after warm motions are running, with daily approval batches of 10 free/basic or 20 Sales Navigator/Premium contacts.\n\n---\n\n## The plays that drive the most value\n\n1. **Social Engagement Flow** - act on post likes/comments first; for founders this is the number one starter play and can use owned posts or a relevant competitor founder/influencer profile.\n2. **Social Campaigns** - build a narrow campaign using either row-level dynamic approvals for rep/BDR pushes or static templates with flow-level approval for governed one-time campaigns.\n3. **AI SDR / ICP Outbound** - build daily approval-ready batches from a HubSpot list or FirstTouch Discover Contacts.\n4. **Founder AI SDR** - the AI SDR workflow in founder voice, with the same approval queue and recommended 10/20 daily connection caps.\n5. **Inbound Speed-to-Lead** - attach LinkedIn connection and follow-up to signups, trials, or demo requests.\n6. **Website Visitor Follow-Up** - act on pricing/demo/product-page intent from HubSpot tracking or RB2B/list sources.\n7. **HubSpot Signal Touches** - turn lifecycle, list, or deal events into timely social touches.\n8. **Stalled Deal Reactivation** - trigger a contact-based HubSpot workflow for contacts associated to open deals not Closed Won/Lost with no engagement for 60+ days, then draft a fresh approved LinkedIn touch.\n9. **Customer Referral Thank-You** - connect with new customers, thank them, ask for feedback, and invite light referrals.\n10. **Sequence QA / Workspace Audit** - make sure the setup and messaging are safe before scaling.\n\n---\n\n## Plays at a glance\n\n| Play | Outcome | Needs / fallback |\n|------|---------|-------|\n| FirstTouch Messaging | Draft on-brand, high-converting LinkedIn outreach | No HubSpot required |\n| Warm Engager Follow-Up | Turn post engagers into conversations | FirstTouch; HubSpot optional for qualification/routing |\n| Social Campaigns | Build one-time LinkedIn campaigns for narrow ICP segments with static templates and flow-level approval | No HubSpot required for Discover Contacts/imported lists; HubSpot required for CRM/deal/customer segments |\n| AI SDR / ICP Outbound Builder | Pull target accounts, qualify, find personas, and add to a daily approval queue | HubSpot list preferred; otherwise ICP brief + FirstTouch Discover Contacts |\n| Founder-Led AI SDR | Run the AI SDR workflow in founder voice | HubSpot list preferred; otherwise ICP brief + FirstTouch Discover Contacts |\n| Inbound Speed-to-Lead | Attach LinkedIn connection + follow-up to inbound signups or trials | HubSpot or FirstTouch-accessible inbound list/import |\n| HubSpot Signal → LinkedIn Touch | Act on CRM lifecycle/list/deal signals with social outreach | HubSpot required |\n| Website Visitor Follow-Up | Turn website visitor intent into social outreach | HubSpot tracking or RB2B/list source |\n| Stalled Deal Reactivation | Contact-based HubSpot workflow for contacts associated to open deals not Closed Won/Lost with no engagement for 60+ days, then queue reactivation touches | HubSpot required |\n| Customer Referral Thank-You | Connect with new customers, thank them, ask for feedback, and invite light referrals | HubSpot Closed Won/customer source or imported customer list required |\n| HubSpot Social Task Runner | Execute due HubSpot LinkedIn/social tasks through FirstTouch and mark them complete after queueing | HubSpot MCP with task read/write required |\n| Team Performance Report | Summarize sends, replies, meetings, opportunities, and HubSpot logging coverage by flow/sender/date | FirstTouch; HubSpot improves logging coverage checks |\n| Sequence QA Reviewer | Review campaigns for risk + quality | No HubSpot required |\n| Campaign Pause & Fix | Pause a live flow mid-send, diagnose, fix copy/audience/volume, and restart with safe re-enrollment | No HubSpot required |\n| Founder Post Generator | Draft founder-voice posts that attract ICP engagers; feeds the warm-engager play | No HubSpot required; drafts only |\n| Post-Demo Follow-Up | Same-day buyer follow-up, champion thank-you, stakeholder expansion, momentum touch | HubSpot recommended; works from meeting notes |\n| Champion Job Change | Congrats now, reconnect later, protect the old account when a champion moves | HubSpot recommended for CRM updates |\n| Workspace Audit | Find setup gaps before launch | No HubSpot required; HubSpot improves CRM checks |\n\n---\n\n## Design principles (Anthropic Agent Skills standard)\n\n- **Progressive disclosure** - only the relevant context loads. Metadata is always present; the body loads on trigger; references load on demand.\n- **Passive, deliverable-oriented** - each skill defines a deliverable, not a goal. The agent produces an artifact; a human approves before anything sends.\n- **Human-in-the-loop by default** - the agent never sends or publishes a flow without chat approval. Dynamic plays require per-send approval; one-time social campaigns can use flow-level approval for the exact audience and static templates. FirstTouch also offers an optional in-product approval layer (off by default) that pauses sends as approval tasks in FirstTouch or HubSpot - recommended for any flow that keeps running after the chat ends (see [`references/safety-governance.md`](references/safety-governance.md)).\n- **HubSpot-aware, not HubSpot-blocked where possible** - HubSpot-specific plays say so clearly; AI SDR and founder AI SDR can run from FirstTouch Discover Contacts when no HubSpot list exists.\n\n---\n\n## Persona Packs\n\nIn addition to the flat `skills/` folder above, this repo ships **4 downloadable persona packs** - self-contained bundles that collect the right skills for a specific role.\n\n| Pack | Persona | Skills included | Key plays |\n|---|---|---:|---|\n| `founder-pack` | Founders doing their own sales | 11 | Social engagement flow, social campaigns, founder AI SDR, inbound follow-up, customer referral thank-you, stalled open-deal workflow |\n| `ae-pack` | Account Executives | 13 | Inbound speed-to-lead, social campaigns, customer referral thank-you, stalled open-deal workflow, meeting-booked stakeholder follow-up, closed-lost re-engage |\n| `bdr-pack` | Business Development Reps | 9 | Inbound speed-to-lead, social engagement flow, social campaigns, target-list AI SDR, lead recovery |\n| `revops-pack` | Revenue Operations | 13 | Workspace audit, sequence QA, customer referral thank-you, social campaigns, stalled open-deal workflow, HubSpot workflow build, team governance |\n\n### Shared-core model\n\nAll 4 packs include the same 6 core plays, each with a persona-specific lens:\n\n1. **Inbound speed-to-lead** - fast LinkedIn touch on signups, trials, demo requests\n2. **Social engagement flow** - act on post engagement from your posts, leadership posts, or company content\n3. **Social campaigns** - one-time campaigns for narrow ICP segments with static flow templates and flow-level approval\n4. **Website visitor play** - turn high-intent page visits into outreach\n5. **AI SDR build** - daily approval-ready outbound from a HubSpot contact/company list, or from a newly discovered ICP list when no HubSpot list exists\n6. **Scoop-up slipped leads** - recover dormant MQLs, no-shows, and stale pipeline when HubSpot/list data exists. Stalled-deal reactivation specifically means a contact-based HubSpot workflow enrolling contacts associated to open deals not Closed Won/Lost with no engagement for 60+ days.\n\nEach pack then adds persona-unique plays on top of this shared core.\n\n### How packs are built\n\nPacks are **vendored copies** of canonical skills - no symlinks, no shared imports. This is required by the Agent Skills standard: every install path (Claude Code `~/.claude/skills/`, Claude.ai zip upload, Cursor folder copy) works with isolated skill folders. The build script handles sync so canonical `skills/` remains the single source of truth.\n\n```bash\npython scripts/build-packs.py\n```\n\nThis reads each `packs/<persona>.json` manifest and materializes:\n\n```\ndist/\n├── founder-pack/\n│   ├── README.md          ← persona-framed plays + install guide\n│   ├── skills/            ← copies of this persona's skills\n│   └── references/        ← onboarding, MCP setup, safety, messaging\n├── founder-pack.zip       ← ready to upload to Claude.ai\n├── ae-pack/  ...\n└── revops-pack/  ...\n```\n\nRe-running the script syncs all packs to the current canonical state, rebuilds the zips in `dist/`, and publishes the distributable zips to `packs/*-pack.zip` plus self-contained single-skill zips to `packs/skills/`. CI validates frontmatter, manifests, and zip freshness on every push.\n\n---\n\n## Support\n\n- **Slack community** - [firsttouch-workspace.slack.com](https://firsttouch-workspace.slack.com): the best place to ask questions, share plays, and get help from the FirstTouch team.\n- **In-app chat** - chat support inside the FirstTouch app for account-specific issues.\n- **Something broke mid-play?** - [`references/troubleshooting.md`](references/troubleshooting.md) covers MCP failures, cooldowns, stalled queues, and credit issues.\n\n*Pack version 1.5.1 - last updated 2026-09-21. See [CHANGELOG.md](CHANGELOG.md).*\n"}