Shorter replies from pr-actions-detective #684

Activity

1. github-actions commented on Mar 13, 2026

   

   github-actionsBot

   on Mar 13, 2026 – with GitHub Actions

   Contributor

   More actions

   3 related issues

   • https://github.com/elastic/integrations/pull/17730 contains multiple comments for the PR Buildkite Detective #647 — It focuses on reducing detective comment noise by updating/reusing one comment, which is useful context for making replies less verbose.
   • Combine multiple automated comments into one with collapsable sections #649 — It proposes consolidating automated outputs into a single collapsible comment, which directly helps keep replies shorter and easier to scan.
   • Noop on no failure #605 — It tracks earlier behavior in the same detective flow and linked external PR context, which helps explain prior verbosity/no-op adjustments.

   ---

   What is this? | From workflow: Trigger Duplicate Issue Detector

   Give us feedback! React with 🚀 if perfect, 👍 if helpful, 👎 if not.
2. github-actions commented on Mar 13, 2026

   

   github-actionsBot

   on Mar 13, 2026 – with GitHub Actions

   Contributor

   More actions

   tl;dr: pr-actions-detective is already deduping/no-oping correctly for no-failure cases; the remaining gap for this issue is response verbosity, which is currently driven by a rigid required comment template. I recommend tightening the prompt format with explicit brevity budgets.

   Recommendation

   Update .github/workflows/gh-aw-pr-actions-detective.md Step 3 to require a minimal response shape with hard limits (for example: TL;DR ≤ 1 sentence, Remediation ≤ 3 bullets, details ≤ 5 bullets total, log excerpt ≤ 8 lines), then run make compile so lock files inherit the change.

   Findings

   • The detective prompt currently mandates a multi-section structure and says “Use this exact shape,” which encourages long comments:
     • .github/workflows/gh-aw-pr-actions-detective.md:132-166
     • Includes required TL;DR, expanded Remediation, and detailed sections (Root Cause, Evidence, Validation, Follow-up).
   • Safe output controls comment count/visibility, not brevity:
     • .github/workflows/gh-aw-fragments/safe-output-add-comment-pr-hide-older.md:3-9 (max: 1, hide-older-comments: true)
     • .github/workflows/gh-aw-fragments/safe-output-add-comment-pr-hide-older.md:14 (body max 65,536 chars)
     • .github/workflows/gh-aw-pr-actions-detective.lock.yml:761 ("maxLength": 65000)
   • Related no-failure/noise work appears already addressed and is distinct from this issue’s ask:
     • #605 (open), #620 (closed), #644 (closed), #609 (merged), #596 (merged).

   Verification

   I verified the current behavior/instructions with:

   nl -ba .github/workflows/gh-aw-pr-actions-detective.md | sed -n '132,166p'
   nl -ba .github/workflows/gh-aw-fragments/safe-output-add-comment-pr-hide-older.md | sed -n '1,22p'
   nl -ba .github/workflows/gh-aw-pr-actions-fixer.md | sed -n '136,143p'
   grep -nE 'maxLength|hide_older_comments|max":1|Maximum 1 comment' \
     .github/workflows/gh-aw-pr-actions-detective.lock.yml | head -n 20

   Observed output confirms:

   • fixed verbose response template in detective prompt,
   • no runtime short-length guardrail beyond very high max length,
   • other workflows use shorter instruction style (e.g. gh-aw-pr-actions-fixer.md:138-142).

   Detailed Action Plan

   1. Edit .github/workflows/gh-aw-pr-actions-detective.md Step 3:
      • Replace “Use this exact shape” with “Use the minimal shape below; omit empty sections.”
      • Add explicit length/bullet budgets.
   2. Keep the existing top-level structure (TL;DR, Remediation, optional <details>) so downstream readers still get consistent navigation.
   3. Recompile generated workflows:
      • make compile
   4. Validate:
      • Trigger on one failed PR check and confirm the posted detective comment is materially shorter but still actionable.
      • Confirm no regressions in safe-outputs behavior (max: 1, hide older comments still active).
   5. (Optional hardening) If you want enforcement beyond prompt-following, add a future compiler/runtime feature for per-workflow response-length caps.

   Related Items

   Type	Link	Relevance
   Issue	#684	Current request: shorter detective replies
   Issue	#605	No-failure noop history (open)
   Issue	#620	No-failure noop request (closed)
   Issue	#644	Similar no-failure noise report (closed)
   Issue	#649	Broader multi-comment noise discussion
   PR	#596	Enabled hide-older-comments for detective workflows
   PR	#609	Merged no-failure noop updates
   File	.github/workflows/gh-aw-pr-actions-detective.md:132-166	Current verbose response template
   File	.github/workflows/gh-aw-fragments/safe-output-add-comment-pr-hide-older.md:3-9	Count/visibility controls
   File	.github/workflows/gh-aw-pr-actions-detective.lock.yml:761	High max body length in generated schema

   ---

   What is this? | From workflow: Trigger Issue Triage

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