# Grok Bot Collector > Searches X (Twitter) for new Grok Bot announcements and adds them to the grokbots directory. Use when the user asks to collect bots, search X for bots, find new grok bots, update the bot directory, or scrape bot announcements from X posts. - Skill: `linxiaotao/grok-bot-collector` (Agent Skill) - Install (CLI): `npx skillmds@latest add linxiaotao/grok-bot-collector` - Raw SKILL.md: https://api.skillmd.com/api/skills/linxiaotao/grok-bot-collector/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: linxiaotao (https://skillmd.com/u/linxiaotao) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/linxiaotao/grok-bot-collector --- # Grok Bot Collector Collects Grok Bot entries from X posts via AgentEarth MCP and appends new bots to the project's bot data files. ## Prerequisites - AgentEarth MCP namespace `user-agentearth` must be available. - Bot data lives in `src/data/bots/` (split by category, barrel-exported from `src/data/bots/index.ts`) ## Workflow Checklist When the user says "search X for new grok bots", "collect new bots from X", or similar: - [ ] **Step 1** — Search X for Grok Bot posts - [ ] **Step 2** — Parse results into bot entries - [ ] **Step 3** — Check for duplicates against existing bots - [ ] **Step 4** — Add new bots to the appropriate category file - [ ] **Step 5** — Sync `README.md` and `README.zh.md` per `.cursor/rules/sync-readme-on-bot-update.mdc` - [ ] **Step 6** — Run `pnpm build` to verify Report a summary at the end: posts searched, bots found, bots added, duplicates skipped. --- ## Step 1: Search X for Grok Bot posts ### 1a. Discover the X search tool Call `RecommendTools` in namespace `user-agentearth`: ``` query: "search X Twitter posts" ``` Pick the search tool from results. Prefer: | Tool | Use when | |------|----------| | `ae_x_twitter_searchpostsrecent` | Last 7 days (5 credits/call) | | `ae_x_twitter_searchpostsall` | 8–30 days ago via `start_time` (5 credits/call) | Use `GetToolDetail` if the input schema is unclear. ### 1b. Run searches Call `ExecuteTool` with the exact `tool_url` from `RecommendTools` (do not modify it). Run multiple queries to maximize coverage: ``` "grok bot" -is:retweet "Grok Bot" -is:retweet "Grok Bot template" -is:retweet "@bot is ready" grok -is:retweet "Meet my Bot" grok -is:retweet ``` Suggested params for each search: ```json { "query": "(\"grok bot\" OR \"Grok Bot\") -is:retweet lang:en", "max_results": 25, "sort_order": "recency", "expansions": "author_id", "tweet.fields": "created_at,text,author_id,public_metrics", "user.fields": "name,username" } ``` **30-day window:** Compute `start_time` as 30 days before today in UTC ISO 8601 (e.g. `2026-07-31T00:00:00Z`). Use `ae_x_twitter_searchpostsall` with `start_time` for the full range; use `searchpostsrecent` for the freshest 7 days. Paginate with `next_token` only when results look promising and credits allow. Tell the user estimated credit cost before running many queries. ### 1c. Filter results Keep posts that look like **bot announcements**, not generic Grok chat: - Mentions creating, sharing, or launching a Grok Bot - Contains a bot name and what it does - Includes phrases like "my bot", "built a bot", "bot template", "Meet my Bot", "is ready" Skip: replies with no bot info, pure hype, retweets, unrelated Grok product news. - Must be possible to find or derive a `https://x.ai/bot/{id}` URL for the bot --- ## Step 2: Parse results **`xaiBotUrl` is required.** If the collector cannot find a valid `https://x.ai/bot/{id}` URL for a bot, skip that bot and list it in the summary as "needs manual xaiBotUrl". For each qualifying post, extract: | Field | Source | |-------|--------| | **name** | Bot title in post (often quoted, Title Case, or after "Meet my … Bot") | | **author** | Display name from expanded user object | | **authorHandle** | `@username` from author | | **description** | What the bot does — 1–2 sentences, rewrite clearly if needed | | **integrations** | Services mentioned: Gmail, Google Calendar, Slack, GitHub, LinkedIn, etc. Use `[]` if none | | **xPostUrl** | `https://x.com/{username}/status/{tweet_id}` | | **xaiBotUrl** | `https://x.ai/bot/{id}` — extract from post text, linked URLs, or by visiting the bot page on x.ai | | **category** | Infer from description (see table below) | | **slug** | Lowercase name → hyphens, strip special chars (e.g. `"Receipt Scanner"` → `receipt-scanner`) | ### Category heuristics | Category | Signals | |----------|---------| | `engineering` | code, developer, git, IDE, harness, deploy, API, compiler, SQL | | `sales` | sales, prospect, outreach, deal, pipeline, crm, pitch deck, gtm | | `marketing` | marketing, social media, X, LinkedIn, copywriter, seo, audience, clips | | `design` | design, figma, ui, ux, critique, sprite, palette, art director | | `product` | product idea, competitor, user research, roadmap, founder, review | | `operations` | workflow, office ops, facilities, admin, notion, legal, audit, invoices, expense | | `recruiting-people` | recruit, talent, hiring, interview, ats, candidate, people | | `personal` | lifestyle, health, wellness, habits, daily newspaper, credit cards, home robots | Default to `operations` when unclear. ### Bot entry defaults ```typescript { id: "", slug: "", name: "", author: "", authorHandle: "@", description: "<1-2 sentences>", integrations: ["Gmail"], installs: 0, category: "", shape: "teardrop", // optional: blob, pebble, bean, egg, squircle, tablet, capsule, cylinder, hex, gem, crystal, wedge, shield, dome, arch, cloud, teardrop, leaf color: "blue", // optional: red, orange, yellow, green, cyan, blue, violet, magenta, gray, brown, black iconColor: "bg-blue-500", createdAt: "", xPostUrl: "https://x.com/...", xaiBotUrl: "https://x.ai/bot/...", } ``` Only include `longDescription` if the post has substantial extra detail worth preserving. --- ## Step 3: Check for duplicates Read all category files in `src/data/bots/` before adding anything. Skip a candidate if **any** match: - Same `slug` (case-insensitive) - Same `name` (case-insensitive, ignoring minor punctuation) - Same `xPostUrl` - Same `authorHandle` + very similar name Log skipped duplicates in the final summary. --- ## Step 4: Add new bots ### 4a. Assign IDs Find the highest numeric `id` across all category files, then increment for each new bot. ### 4b. Pick iconColor Choose a Tailwind class not already heavily used. Rotate through: `bg-blue-500`, `bg-purple-500`, `bg-green-500`, `bg-orange-500`, `bg-teal-500`, `bg-red-500`, `bg-indigo-500`, `bg-yellow-500`, `bg-emerald-500`, `bg-cyan-500`, `bg-violet-500`, `bg-amber-500`, `bg-rose-500`, `bg-sky-500`, `bg-fuchsia-500`, `bg-pink-500`, `bg-lime-500` ### 4c. Set createdAt Use today's date in `YYYY-MM-DD` format. ### 4d. Append to the correct category file Add new bots to their matching category file in `src/data/bots/.ts`: - Insert new objects before the closing `];` - Add a comma after the previous last entry - Use double quotes for strings, trailing commas on object fields Example entry: ```typescript { id: "80", slug: "weekly-digest-bot", name: "Weekly Digest Bot", author: "Jane Doe", authorHandle: "@janedoe", description: "Summarizes your unread emails every Monday and sends a one-page digest.", integrations: ["Gmail"], installs: 0, category: "operations", iconColor: "bg-cyan-500", createdAt: "2026-08-30", xPostUrl: "https://x.com/janedoe/status/1234567890", xaiBotUrl: "https://x.ai/bot/abc123", }, ``` Do not reorder or modify existing entries unless the user asks. --- ## Step 5: Sync READMEs Follow `.cursor/rules/sync-readme-on-bot-update.mdc`: - Add new entries to `README.md` and `README.zh.md` under their matching category section. - Format: `- **[Bot Name](xaiBotUrl)** by @handle — Description.` --- ## Step 6: Verify ```bash pnpm build ``` If the build fails, fix TypeScript errors and rebuild. --- ## AgentEarth MCP reference ``` namespace: user-agentearth RecommendTools → find X tools by natural-language query GetToolDetail → inspect a tool's input_schema ExecuteTool → run a tool - tool_url: exact URL from RecommendTools (never modify) - params: per input_schema ``` Optional follow-up tools: - `ae_x_twitter_getpostsbyid` — fetch full text of a promising post - `ae_x_twitter_getusersbyusername` — resolve author display name --- ## Edge cases - **Ambiguous post:** Skip rather than guess. Mention in summary. - **Bot name missing:** Derive from first line or skip. - **No integrations mentioned:** Use `[]`. - **Non-English posts:** Skip unless user requests otherwise. - **Credit limits:** Run fewer queries, report partial results. - **Empty results:** Try broader queries (`grok bots`, `"built with grok"`) before giving up.