[agent-suggestions] Agent suggestions - 2026-03-02

[github-actions]

Agent Suggestions

Date: 2026-03-02

1. Actions Failure Cluster Detector

Trigger: Scheduled daily (cron) + workflow_dispatch

Purpose: Detect recurring failure signatures across trigger-* / gh-aw-* workflow runs in a rolling window (for example checkout ref failures, missing agent-output artifacts, runtime/config errors), then file one consolidated, prioritized tracking issue with grouped evidence and suggested remediation owners.

Proposed safe outputs: create-issue (single consolidated report issue per run window; noop when no recurring cluster is found)

Evidence:

• Recurring failure clusters already documented manually, including repeated signatures across many runs/workflows: [agent-efficiency] Recurring agent workflow failures: checkout ref errors and missing agent-output artifact #419
• Recent workflow-failure operational issue requiring manual triage routing: [agentics] Internal: Agent Deep Dive failed #387
• Ongoing reliability hardening work in open workflow PRs:
  • Add merge-conflict handling to mention-in-pr-no-sandbox prompt #514
  • configurable github token for safe outputs #392
  • Rename estc-downstream-health to internal-downstream-health #515

Why not covered already: Existing workflows are PR-scoped or narrow detectors (for example pr-actions-detective, branch-actions-detective, and specific error detectors), but there is no generalized scheduled workflow that clusters recurring failures across runs and continuously emits one consolidated reliability backlog item.

Duplicate Checks

• Open workflow/automation issues reviewed: https://github.com/elastic/ai-github-actions/issues?q=is%3Aissue+is%3Aopen+%28agent+OR+workflow+OR+automation%29
• Open workflow PRs reviewed: https://github.com/elastic/ai-github-actions/pulls?q=is%3Apr+is%3Aopen+%28agent+OR+workflow%29
• Prior report-title check: https://github.com/elastic/ai-github-actions/issues?q=is%3Aissue+in%3Atitle+%22%5Bagent-suggestions%5D%22

Downstream Signals

• Workflow usage/activity signals in downstream repos:
  • https://github.com/elastic/beats/blob/main/.github/workflows/trigger-pr-review.yml
  • https://github.com/elastic/integrations/blob/main/.github/workflows/pr-buildkite-detective.yml
  • https://github.com/elastic/ai-github-actions-playground/blob/main/.github/workflows/flaky-test-investigator.yml

Suggested Next Steps

• Add gh-aw-actions-failure-cluster-detector as a scheduled-audit style workflow with configurable lookback window and workflow filters.
• Emit grouped findings by failure fingerprint and include top-N reproducible examples per cluster.
• Add dedupe logic so repeated daily reports update/append to a single active tracking issue instead of opening noisy duplicates.

---

What is this? | From workflow: Agent Suggestions

Give us feedback! React with 🚀 if perfect, 👍 if helpful, 👎 if not.

• expires on Mar 9, 2026, 12:29 PM UTC

[github-actions]

This issue was automatically closed because it expired on 2026-03-09T12:29:49.939Z.

Closed by Workflow