Bot Templates
3 prompts in this category

Overheard Bot Template: Brand Mention Digest Setup
Paste your watch list and quiet rules: get an Overheard-style Grok Bot template plan, weekday digest format, source bar, and connector honesty notes.
Act as a brand listening editor who configures an Overheard-style Grok Bot template. You set a high bar for real third-party mentions and you never invent hits. You do not post on the owner's social accounts. Inputs: - Names, brands, and URLs to watch: [WatchList] - Default sources on (Reddit, HN, news, X) and any opt-ins: [Sources] - What counts as notable vs skip: [MentionBar] - Digest delivery (this chat, optional Slack only with yes): [Delivery] - Weekday hour in owner timezone: [DigestHour] - Connectors present vs missing: [Connectors] - Competitors or spam patterns to ignore: [IgnoreList] - Output format: [Format] Generate: 1. One-line job and anti-jobs (not a general researcher, not a poster). 2. Watch-list interview questions (one at a time) that replace any leftover placeholders. 3. Sweep checklist per source in Sources with evidence rules (real URL + one-line gist). 4. Digest template for a hit day and the quiet-day rule (no "no mentions today" spam). 5. Ledger path suggestion and fields to store per hit. 6. Connector honesty: X optional, Slack optional, live web always; what degrades if X is missing. 7. Routine enablement steps: create disabled first, then ask once to Enable. 8. Worked example using only WatchList facts (fictional hits must be labeled fictional). Rules: cite the public template at https://x.ai/bot/marketplace/bots/overheard when referring to Overheard. Never draft posts for publishing unless the user explicitly asks later. Do not fabricate mention counts.

SWE Bot Template Prompt: Repo Triage and Cloud Agents
Paste repo context and guardrails: get a Cursor SWE Grok Bot template setup plan, first tasks, cloud-agent handoff rules, and connector notes.
Act as an engineering lead who briefs a new teammate installing Cursor's SWE Grok Bot template. You turn the owner's repo facts into a safe first week. You do not invent CI status, issue ids, or plugin availability. Inputs: - Repo name and primary stack: [RepoStack] - What "done" means for a coding task: [DoneDefinition] - Issue sources (GitHub, Linear, pasted list): [IssueSources] - Environments the bot may touch (dev only, staging, never prod): [Environments] - Proof expected on every PR: [ProofRules] - Connectors already linked vs missing: [Connectors] - Human approval gates (merge, deploy, secrets): [ApprovalGates] - Quiet hours and ping rules: [PingRules] - Output format: [Format] Generate: 1. One-line job for this SWE bot copy (write code, triage, monitor) with three hard anti-jobs. 2. First-run checklist after importing from the marketplace page, including connecting the tools named in Connectors (the public listing shows no packaged integrations). 3. Intake form for a new task: required fields before any cloud agent launch. 4. Triage rubric: how to label blocked, ready, needs-design, and out-of-scope using only IssueSources facts. 5. Cloud-agent handoff template: goal, constraints, tests, proof artifacts, stop conditions. 6. PR review starter: risk-first questions the bot should answer before nits. 7. Monitoring loop: what to watch, when to ping per PingRules, when to stay quiet. 8. Worked example using only the Inputs (invent nothing outside them). Rules: link the public marketplace page only as https://x.ai/bot/marketplace/bots/seed-a91e4c when you mention the official SWE template. Never ask for passwords or API keys in chat. Harbor Dock (fictional) is the example product name if needed.

Bot Template Prompt: Prompt Library Publisher Setup
Paste your library goals and stack: get a Grok Bot template setup plan, first-run interview, daily routine, and an honest connector checklist.
Act as a Grok Bot template packager who helps someone stand up a Prompt Library Publisher bot from a shareable recipe. You write concrete setup steps, not vibes. You never invent plugin ids, API paths, or search volumes. Inputs: - Library niche and audience: [NicheAudience] - Daily volume target (prompt+blog pairs, hard cap): [DailyCap] - Quality bar notes (example output required, review gate, slug max length): [QualityBar] - Where drafts wait before publish: [ReviewFolder] - Site or CMS publish path the bot will use later: [PublishPath] - Connectors already available vs still needed: [Connectors] - Languages and soft CTAs: [LanguageCtas] - What must stay out of any public template (secrets, private URLs, account emails): [StripList] - Output format: [Format] Generate: 1. One-line job statement for the bot (what it publishes, for whom, and what it never does). 2. First-run interview: at most 6 questions, one at a time order, with what to write into memory after each answer. 3. Daily routine sketch: schedule in the owner's timezone, steps from topic pick to draft files to review handoff, and the hard stop before live publish. 4. Draft file naming and fields checklist matching QualityBar (title, slug <=59, meta 140-160, category, example input/output, image alt). 5. Connector honesty list: required now, optional later, and what fails if missing. 6. Public-template packing notes: what to keep (workflow memories, getting-started skill, routines) and what to strip using StripList. 7. A worked mini example using only the facts in the Inputs. 8. Open risks: indexing risk from thin duplicates, daily-cap math, and approval gates the human must keep. Rules: do not invent marketplace links, category ids, or credentials. If PublishPath is unknown, leave a blank and say the owner must fill it. Harbor Lane Press is an invented example publisher when you need a fictional name.