# Community Radar > Monitors public community or social mentions of the brand, competitors, or relevant pain language. Classifies sentiment and drafts short, human replies for the right responder. - **Type:** Skill - **Install:** `agentstack add skill-swan-gtm-gtm-skills-community-radar` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [swan-gtm](https://agentstack.voostack.com/s/swan-gtm) - **Installs:** 0 - **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [swan-gtm](https://github.com/swan-gtm) - **Source:** https://github.com/swan-gtm/gtm-skills/tree/main/skills/ido-goldberg/community-radar ## Install ```sh agentstack add skill-swan-gtm-gtm-skills-community-radar ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About ## Instructions **Setup state.** Not yet configured for this org. Load the `Setup` sub-page and walk the user through wiring the community-radar trigger (`SCHEDULE` + Apify search actors across LinkedIn / X / Reddit / Hacker News, or `WEBHOOK` from a social listening tool the user already pays for) with the org's brand-and-pain-language keyword list, the platforms to monitor, and the follow-up sequence before running this play. (After setup is performed, rewrite this paragraph via `swan-update-skill` to describe the current state — trigger type chosen, keywords / brand terms / pain-language phrases monitored, platforms covered, sequence wired, success metric, and last-refreshed date — so future runs see the current configuration without re-checking.) ### When this fires A `SCHEDULE` trigger running an Apify search actor (LinkedIn search, X / Twitter search, Reddit search, Hacker News scrape) surfaces new mentions of the configured keywords. Or a `WEBHOOK` from a social listening service the user already pays for (Brand24, Mention, Hootsuite, Triggify, custom Reddit / X monitoring) pushes mentions in. Payload includes: platform, author, mention text, post URL, engagement metrics on the parent post. Note: `LINKEDIN_ENGAGEMENT` is *not* the right trigger here — that one follows specific LinkedIn profiles. For keyword-driven brand-mention sweeps across LinkedIn, use `SCHEDULE` + an Apify LinkedIn-search actor. The window is short on public social — 24-48 hours feels reasonable; > 1 week and the reply looks bot-driven. ### Step 1 — Classify the mention | Class | Pattern | Right move | |-------|---------|------------| | **Direct praise** | "We love [your product]" | Like, optional thank-you reply. Resharable. | | **Customer Q / mild frustration** | "How do I do X in [your product]?" | Helpful reply from support handle. Resolve the question. | | **Public complaint** | "[Your product] is broken / disappointing" | Acknowledge, DM to take offline, don't argue publicly. | | **Comparison shopping** | "Looking at [you] vs [competitor]" | Soft entry; offer to help with the eval. Don't trash competitor. | | **Competitor switch signal** | "Just switched off [competitor]" + same thread mentions you | High-value lead; warm DM. | | **Pain mention (your wedge, no brand)** | "Why can't I find a tool for X" | Soft helpful reply. Don't pitch — offer perspective. | | **Generic noise / spam / off-topic** | — | Ignore. | ### Step 2 — Identify the author `swan-fetch-scraped-url` on the author's profile (LinkedIn, Twitter bio, Reddit profile). Capture: role, company, follower count, post pattern. Don't enrich if it's clearly noise. For LinkedIn: `swan-enrich-contact` if they look ICP-fit. For other platforms: company affiliation is often in bio; cross-check via `swan-search-companies`. ### Step 3 — ICP and CRM context For mentions from ICP-fit authors: - `swan-search-companies` + (if new) `swan-enrich-company` - `hubspot-search-objects` for existing relationship For non-ICP: still respond if it's a complaint or Q (support obligation), but don't pursue. ### Step 4 — Choose the right responder Public replies should come from the right account: | Class | Right responder | |-------|-----------------| | Praise | Founder / CEO (high-status reply) | | Customer Q | Support handle / CSM | | Complaint | Support handle, then CSM via DM | | Comparison shopping | AE, via DM not public comment | | Switch signal | AE, via DM, fast | | Pain mention | Founder / thought leader, public comment | If multiple senders are connected, pick the one whose voice fits the moment. Don't auto-reply from a generic brand account if a person's voice would land better. ### Step 5 — Draft the reply Templates: **Public complaint:** > "Sorry to hear this. DMing now — want to get this sorted today." > (Then DM with substance and a fix.) **Comparison shopping (DM):** > "Saw your post — happy to help with the eval, no pitch. What matters most for you in [category]? I can be straight about where we win and where we don't." **Pain mention (no brand):** > "Same — this is one of those problems that's worse than people say. Our take: [one-line perspective]. Happy to share more if useful." **Switch signal (DM):** > "Just saw your post about leaving [competitor] — congrats on the cleanup. If [your product] is on the eval list, glad to give you the no-pitch tour." Critical: short, human, no marketing. Public social rewards low-key over polished. ### Step 6 — Channel: public reply vs DM Default to DM for anything sales-adjacent. Public replies should be ones you're OK with anyone in the future reading — they live forever and get screenshot. Public is right for: praise threads (you're amplifying), pain mentions where helpful insight beats outreach, supportive Q&A. ### Step 7 — Route or send For LOW-stakes (praise, generic Q): the system can auto-reply with the right account. Surface for approval if voice matters. For MEDIUM-stakes (comparison shopping, switch signals): hand off to the AE via `hubspot-create-task` — let the human draft. Sales DMs need human nuance. For HIGH-stakes (complaint, brand crisis): notify the right responder via `slack-send-notification` immediately. Don't let a slow CRM task be the bottleneck. ### Step 8 — Log `swan-update-company` to log the mention. If a complaint, log both the issue and the resolution path. If a switch signal converts, log the source — social mentions that convert are some of the highest-ROI to track over time. ### Rules - MUST classify the mention before drafting. The reply for praise and the reply for complaint are different jobs. - MUST keep public replies human and low-key. Polished marketing in a Reddit thread is brand suicide. - MUST DM-not-public for anything sales-adjacent. - NEVER argue publicly with a complaint. Acknowledge, take it offline, fix it. - NEVER name a competitor pejoratively in a public reply. Even if the OP did. - NEVER auto-reply to high-stakes mentions. Humans only. - If sentiment is escalating (multiple replies, growing engagement on a negative thread), escalate fast — that's a brand crisis, not a routine signal. - If a tool result is truncated, read from `files/tool-outputs/_.json` in `swan-execute-code`. ### Tighten over time After 10-20 fires, read the responder log via `swan-search-sequences` and review which mention classes actually converted (switch signals and comparison shopping usually outperform pain mentions). Drop low-yield keywords from the trigger, tighten the keyword list to brand + competitor + 3-5 highest-signal pain phrases, and revisit the platform mix — Reddit often outperforms X for B2B signal. GAP: native social listening (beyond LinkedIn) isn't a Swan-native tool today. Customers route X / Reddit / Hacker News mentions via webhook from external listening tools. ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [swan-gtm](https://github.com/swan-gtm) - **Source:** [swan-gtm/gtm-skills](https://github.com/swan-gtm/gtm-skills) - **License:** MIT Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-swan-gtm-gtm-skills-community-radar - Seller: https://agentstack.voostack.com/s/swan-gtm - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.