# Heartbeat Examples Use these as inspiration during the interview. Adapt to the user's tools, APIs, and delivery preferences. **Key constraint:** Murmur is just a scheduler — it runs Claude with a prompt and classifies the output. It cannot send notifications on its own. If the user wants results delivered somewhere (Slack, Telegram, a file, a database), the HEARTBEAT.md must include that step explicitly. The heartbeat prompt is the entire action — from gathering data to delivering output. ## Delivery Patterns Every heartbeat that produces output needs a destination. Common patterns: - **Slack webhook**: `curl -X POST -H 'Content-Type: application/json' -d '{"text":"..."}' $SLACK_WEBHOOK_URL` - **Telegram bot**: `curl -s "https://api.telegram.org/bot$TELEGRAM_TOKEN/sendMessage" -d "chat_id=$CHAT_ID&text=..."` - **Append to local file**: Write findings to a markdown file in the workspace - **GitHub issue/comment**: Use `gh issue create` or `gh issue comment` - **Ntfy push notification**: `curl -d "message" ntfy.sh/your-topic` - **ATTENTION response**: Murmur logs it — useful if the user checks `murmur status` or the TUI Remind the user: if they want to be _notified_, the heartbeat itself must do the notifying. Murmur won't forward anything. --- ## Code & Repos ### Codex: architecture review ```markdown --- agent: codex sandbox: workspace-write cron: "0 6 * * 1" timeout: 15m --- You are Martin Fowler — the author of "Refactoring" and "Patterns of Enterprise Application Architecture." You've been asked to review this codebase as a personal favor. Read every source file, trace the dependency graph, and evaluate the architecture holistically. Apply your signature lens: - Smell out code that needs refactoring — duplications, long methods, feature envy - Identify missing or leaky abstractions - Evaluate whether the module boundaries reflect the actual domain - Call out shotgun surgery, divergent change, and inappropriate intimacy - Judge whether the code communicates its intent clearly Write like Fowler: precise, pedagogical, with concrete examples. Don't soften your critique — this team asked for honest feedback. Write your review to `docs/architecture-review.md` with clear sections, specific file references, and concrete recommendations. Rate the overall architecture health on a scale of 1-10. Respond HEARTBEAT_OK when done. ``` ### Auto-triage incoming issues ```markdown --- interval: 30m --- Check for new GitHub issues on {org}/{repo} using `gh issue list --state open --json number,title,body,labels`. For any issue with no labels: - Read the title and body - Apply one of: `bug`, `feature`, `question`, `security` - Use `gh issue edit {number} --add-label {label}` If any issue is labeled `security`, post to Slack: `curl -X POST -H 'Content-Type: application/json' -d '{"text":"Security issue #{number}: {title}"}' {webhook_url}` If no unlabeled issues, respond HEARTBEAT_OK. ``` ### Stale PR nudge ```markdown --- interval: 12h --- List open PRs on {org}/{repo} with `gh pr list --json number,title,author,createdAt,reviewRequests`. For any PR open > 48 hours with no reviews: - Post a comment: `gh pr comment {number} --body "Friendly nudge: this PR has been waiting for review."` - Send a Telegram message: `curl -s "https://api.telegram.org/bot{token}/sendMessage" -d "chat_id={id}&text=PR #{number} needs review: {title}"` If all PRs are reviewed or < 48h old, respond HEARTBEAT_OK. ``` ### CI failure digest ```markdown --- interval: 1h --- Check the last 10 GitHub Actions runs: `gh run list --limit 10 --json status,conclusion,name,headBranch,createdAt`. Collect any with conclusion=failure. If failures exist, write a summary to `ci-report.md` in this workspace: - Workflow name, branch, when it failed - Append to the file (don't overwrite previous entries), date-stamped. If all green, respond HEARTBEAT_OK. ``` ## Research & Intelligence ### Hacker News scout ```markdown --- interval: 6h --- Fetch top 30 HN stories via `curl -s https://hacker-news.firebaseio.com/v0/topstories.json`. For each, fetch details: `curl -s https://hacker-news.firebaseio.com/v0/item/{id}.json`. Filter for stories about: {topics}. If relevant stories exist, append them to `hn-digest.md` in this workspace: - Date header - Each story: title, URL, score, comment count - Sorted by score descending Also post the top 3 to Slack: `curl -X POST -H 'Content-Type: application/json' -d '{"text":"HN digest:\n1. {title} ({score}pts) {url}\n..."}' {webhook_url}` If nothing relevant, respond HEARTBEAT_OK. ``` ### Competitor changelog tracker ```markdown --- interval: 1d --- Fetch the changelog/release pages for: - `curl -s https://github.com/{competitor}/releases.atom` - `curl -s {competitor_changelog_url}` Check for releases or announcements published in the last 24 hours. If anything new, write a summary to `competitive-intel.md` in this workspace and send a Telegram message with a one-liner per release. If nothing new, respond HEARTBEAT_OK. ``` ### Arxiv paper monitor ```markdown --- interval: 1d --- Search recent arxiv papers using the API: `curl -s "http://export.arxiv.org/api/query?search_query=all:{keywords}&sortBy=submittedDate&sortOrder=descending&max_results=10"` For each paper, extract title, authors, abstract, and link. Filter for papers relevant to: {research_focus}. If relevant papers found, append to `papers.md`: - Date header - Title, authors, one-sentence summary, link If nothing relevant, respond HEARTBEAT_OK. ``` ## Ops & Infrastructure ### Endpoint canary ```markdown --- interval: 15m --- Check these endpoints and record response status + latency: - `curl -s -o /dev/null -w "%{http_code} %{time_total}" {url_1}` - `curl -s -o /dev/null -w "%{http_code} %{time_total}" {url_2}` If any returns non-200 or latency > 2s, send a push notification: `curl -d "ALERT: {url} returned {status} ({latency}s)" ntfy.sh/{topic}` Log every check to `uptime.csv` (timestamp, url, status, latency). If all healthy, respond HEARTBEAT_OK. ``` ### Docker resource check ```markdown --- interval: 6h --- Run `docker system df` and `docker ps --format '{{.Names}} {{.Status}}'`. - If disk usage > 80%, run `docker system prune -f` and log what was cleaned. - If any container is in "Restarting" or "Exited" state, report it. Append results to `docker-health.log` in this workspace. If everything is healthy and disk < 80%, respond HEARTBEAT_OK. ``` ## Personal & Creative ### Daily journal prompt ```markdown --- interval: 1d --- Create today's journal entry file: `journal/{YYYY-MM-DD}.md`. Include: - A thoughtful writing prompt based on the current date, season, or recent world events - 3 reflection questions - A quote that fits the theme If today's file already exists, respond HEARTBEAT_OK. ``` ### Repo activity digest ```markdown --- interval: 1d --- For each repo in [{repo_list}]: - `gh api repos/{owner}/{repo}/events --jq '.[0:10]'` - Summarize: pushes, PRs opened/merged/closed, issues opened, releases Write a digest to `weekly-activity.md`: - One section per repo - Only include repos with activity - Overwrite the file each time (it's a rolling snapshot) If zero activity across all repos, respond HEARTBEAT_OK. ``` ## Choosing the Right Schedule Use **interval** for fixed-frequency checks. Use **cron** when you need specific times or weekday-only runs. | Use case | Interval | Cron alternative | | -------------------------------------- | ------------ | ------------------------------------- | | Critical uptime (endpoints, services) | `15m` | `*/15 * * * *` | | Active development (CI, tests, triage) | `30m` – `1h` | `*/30 9-18 * * 1-5` (work hours) | | Reviews & collaboration | `6h` – `12h` | `0 9,17 * * 1-5` (9am + 5pm weekdays) | | Research & intelligence | `6h` – `1d` | `0 8 * * *` (daily at 8am) | | Housekeeping (deps, cleanup, digests) | `1d` | `0 3 * * 0` (Sunday 3am) | Cron heartbeats support an optional `tz` frontmatter field (e.g., `tz: America/New_York`) — defaults to local system timezone.