Engineering workflows

When a watched dependency publishes a new release, determine which monorepo packages import it and notify each package's CODEOWNERS with a tailored summary of what changed…

When a Vercel preview smoke check fails, an AI step reads the failing responses and recent commit diff, writes a likely-cause triage summary.

When a promotion gate blocks a release, an agent pulls the failing Sentry issues, correlates them to the GitHub commits in the release.

Every Monday aggregates the week's flaky-test data from GitHub and Honeycomb, publishes a ranked scorecard to Confluence.

Listens for a Sentry alert that a just-promoted release's crash-free rate has dropped below threshold, immediately re-aliases the previous Vercel deployment.

On a weekly schedule, list every outdated dependency, score each by changelog severity and how many internal modules import it.

Watches your flag-management Postgres table for flags that have sat at 100% rollout past a grace window.

Once a week, summarizes all card, license, and revision changes across your watched HuggingFace models into a single readable digest page in Notion and posts the link to Slack.

On a repeated CI failure, posts a Slack message with retry-or-skip buttons and the LLM's recommendation.

Once a week, collects everything merged into main over the past seven days, has AI write a readable shipped-this-week summary, saves it as a Notion page.

Screenshots key pages on a Vercel preview, compares them against the Figma design baseline.

An autonomous agent intakes a Sentry deprecation, researches the upstream changelog, decides whether to scope it, then orchestrates the migration by opening a Linear epic.

An autonomous agent intakes a Sentry deprecation, researches the upstream changelog, decides whether to scope it, then orchestrates the migration by opening a Linear epic.

An agent inspects a flaky test's recent traces, drafts a likely root-cause hypothesis (timing, ordering, network), and posts it on the quarantine issue with a suggested fix.

Listens for a rollout-completed webhook from your flag platform, waits out a configurable bake period.

On MR open, an agent weighs code ownership, recent file expertise, current queue depth, and availability to recommend the strongest reviewer.

When a test is flagged flaky three times in a week, opens a GitHub PR that marks the test as skipped/quarantined, requests review from the code owner.

Every Monday it aggregates the past week's pull request review metrics, has an LLM write a short narrative on bottlenecks and trends.

When a watched package publishes a new release, fetches the release notes, maps them to the internal modules that depend on it.

Transcribes a shared Loom bug video, classifies its severity and affected area, then creates a Linear issue routed to the right team with priority set.

When a PR fails the license gate, this workflow creates a tracked Linear remediation issue with the offending packages and suggested replacements, links it back to the PR.

Monitors the rate of newly quarantined specs and total CI failure volume; when both spike past a threshold in a short window, it pages the on-call engineer and posts a meltdown…

On a weekly schedule, gathers all of the week's recorded architecture syncs, summarizes the decisions made across them.

An agent investigates a spiking Sentry issue, reads the stack trace and recent commits to the affected files, names the most likely suspect commit and owner.

After a Loom code-review walkthrough is recorded, transcribes it, generates a short digest of action items.

When Dependabot opens a security PR, it pulls the linked advisory severity and, for critical or high CVEs.

On a security advisory for a dependency, checks whether vulnerable modules are actually used in your code, and pages on-call only when a reachable internal module is affected.

Monthly, inventories all production services from GitHub, matches each to a Confluence runbook.

When an MR opens but every matching code owner is over capacity or unavailable, escalates to a fallback reviewer pool and opens a Linear task flagging the ownership gap.

On demand, an agent compares each service's recent GitHub commit history against its Confluence runbook, reasons about which runbook sections are likely outdated.

When a deploy freeze is triggered by budget exhaustion, an agent investigates recent deploys and error patterns, drafts a recovery plan, files a Linear incident issue.

On a daily schedule, re-smoke-tests the live preview for every open pull request and posts a single digest to Slack listing which previews are now broken or stale so reviewers…

On a schedule, queries historical CI test results in BigQuery to rank the most flaky and most-quarantined tests over the quarter.

Once a week, compiles every test currently carrying the flaky-quarantine label into a Notion review board with age, flake history, and owner, and de-quarantines tests that have…

When a test is confirmed flaky three times, an agent opens a GitHub pull request that marks the test as skipped with a tracking comment.

On a daily schedule, pulls per-test flakiness rates from Datadog CI Visibility, quarantines any test above a threshold by filing a labeled Linear ticket.

For dependencies whose license cannot be auto-classified, an agent researches the project's actual licensing online and routes a structured approve/reject request…

On a daily schedule, reads per-test pass/fail metrics from Datadog CI Visibility, and for any test whose intermittent failure rate crosses a threshold.

On a daily schedule, reads per-test pass/fail metrics from Datadog CI Visibility, and for any test whose intermittent failure rate crosses a threshold.

When Sentry marks a resolved issue as regressed, finds the Linear bug previously linked to that fingerprint, reopens it, and notifies the owning team's Slack channel instead…

When a test is newly quarantined, an agent pulls recent failure logs from Datadog and the test source from GitHub, drafts a likely-root-cause analysis.

On a daily schedule, queries Datadog CI Visibility for tests whose flakiness rate exceeds a threshold, quarantines each one in the repo.

When Sentry flags a new high-severity error tagged as security, this drafts a triaged summary with OpenAI, opens a labeled GitLab issue assigned to the security reviewer group.

A Paperclip agent investigates a new Sentry cluster end to end, traces the blame, reads the offending source in GitLab.

An agent audits the codebase for stale feature flags, removes each one branch-by-branch, runs the test suite in a shell sandbox to prove nothing broke.

When a Sentry spike creates a deduplicated Linear bug, an agent reads the stack trace and recent GitHub commits to the culprit file.

An agent crawls the repository for every feature-flag reference, cross-checks each against rollout state, and produces a prioritized removal plan with per-flag risk notes.

Monitors GitLab merge-train pipeline failures, distinguishes flaky failures from real ones via retry history.

When a Sentry issue crosses a spike threshold, this gathers every Intercom conversation tied to the same error and posts a blast-radius summary to Slack listing affected…

When a Sentry release shows a crash-free regression, an LLM summarizes the top new issues driving it and posts an interactive Slack approval request that links straight…

Receives a JUnit results webhook, has the CEO agent classify each flaky failure by likely root cause, and files an owner-assigned ClickUp task with a proposed fix plan.

On a Datadog fast-burn alert webhook, this workflow confirms the SLO is burning faster than the configured multiplier and pages the on-call engineer through PagerDuty…

On a schedule, evaluates aggregate flake load from Datadog and the open deflake backlog in Linear, and if the main pipeline's flake rate threatens releases.

Ask in chat about a slow query and an agent pulls its Honeycomb traces, cross-references recent deploys and PRs.

Transcribes each Loom code-review walkthrough, summarizes the design rationale and decisions.

Listens for Replicate prediction-completed webhooks, flags low-confidence or errored outputs, files each one as a GitHub regression-bench issue with reproduction inputs.

After a design sync is recorded, drafts the ADR in Confluence and opens a tracking pull request stub in GitHub that references the ADR.

Polls Datadog CI Test Visibility on a schedule, finds tests whose flake rate crosses a threshold, and files a Linear quarantine ticket assigned to the last committer.

When Honeycomb signals a critical fast-burn event, pages the on-call engineer via PagerDuty and enables branch protection that blocks merges to the release branch until budget…

Continuously watches a deployed canary's Honeycomb latency and error SLOs and, on a hard breach, reverts the dependency PR to draft, files a PagerDuty incident.