error-pulse

name	error-pulse
description	PostHog alert responder workflow with required Slack deep-links and escalation rules.
publish	false

Error Pulse

Use this skill when responding to PostHog alerts.

Primary goal: close the alert with minimal human toil.

Workflow

1. Investigate root cause first. Gather concrete evidence (logs, traces, failing checks, repro steps).
   • PostHog alert regressions are often in ~/src/github.com/iterate/iterate; check that repo early unless evidence points elsewhere.
2. Classify:
   • False positive / noisy alert: tune threshold, filter expected errors, or suggest silencing with clear rationale. If less logging or narrower logging would safely reduce noise, open a PR.
   • Real bug: implement the smallest safe fix, add tests, and open a PR when the change is trivial, there is an obvious solution, or additional logging would materially improve diagnosis.
3. Keep updates concise and action-oriented.

Autonomous merge policy

You may merge autonomously only when all are true:

1. All required checks are green.
2. Cursor Bugbot labels risk as low (or equivalent low-risk finding).
3. No unresolved review comments remain.
4. Branch is mergeable without conflicts.

If any gate fails, do not merge.

Slack escalation policy

If blocked, uncertain, or impact is broad, post in #error-pulse (C09K1CTN4M7) with evidence and clear asks.

• Use @channel for urgent incidents requiring immediate attention.
• Use @here for important but not time-critical issues.

When you open or continue a Slack thread, subscribe it to your current agent session so replies route back to you:

iterate tool subscribe-slack-thread --channel C09K1CTN4M7 --thread-ts <thread_ts> --session-id <session_id>

Use get-current-session-id tool to get <session_id>.

Slack thread behavior

When you are subscribed to a thread, reply to questions and direct asks even if you are not @mentioned. The #error-pulse channel is an active incident channel - any questions which seem directed at you should be answered, excluding questions where someone else is @tagged.

Slack message requirements

CRITICAL: Every initial post to #error-pulse MUST include deep-links inline. Do NOT post a summary first and add links later - include them in the very first message. This is non-negotiable.

Required deep-links:

• PostHog: link to the error tracking issue or event. The alert payload contains posthogProjectId and $exception_issue_id - use them to build https://<host>/project/<id>/error_tracking/<issue_id>.
• Cloudflare Worker logs: link to the Worker logs page with a pre-filled query for the error message when possible. Use a regex filter on message, URL-encode it, and include it in the very first Slack message. Example shape: https://dash.cloudflare.com/04b3b57291ef2626c6a8daa9d47065a7/workers/services/view/os/production/observability/events?...&filters=%5B%7B%22key%22%3A%22message%22%2C%22operation%22%3A%22regex%22%2C%22type%22%3A%22string%22%2C%22value%22%3A%22<url-encoded-regex-error-message>%22%7D%5D
• Source code: GitHub permalink with line numbers (e.g. https://github.com/iterate/iterate/blob/main/apps/os/src/orpc/root.ts#L12)

For the Cloudflare link, prefer the most specific stable regex you can derive from the alert text. Escape quotes and special regex characters when needed; if the raw error is too noisy, trim it down to the distinctive failure fragment before encoding it into the URL. If you cannot build a reliable pre-filled query, fall back to the normal Cloudflare Worker logs page and say that the filter could not be prefilled confidently.