[PR Triage Report] PR Triage Report — 2026-08-09 (6 PRs)

[github-actions]

PR Triage Report — 2026-08-09

Triaged 6 open PRs authored by app/github-copilot in github/gh-aw.

Executive Summary

Metric	Value
Total triaged	6
Draft PRs	4 (51484, 51473, 51472, 51469)
Ready for review	2 (51465, 51446)
Avg score	46/100
Fast-track candidates	1 (#51446)
Batch-review candidates	1 (#51465)
Defer candidates	4
Auto-merge / close candidates	0

Distribution

• By category: feature (2), refactor (2), docs/chore (1), chore/WIP (1)
• By risk: medium (5), low (1)
• By priority: high (1), medium (1), low (4)
• By action: fast_track (1), batch_review (1), defer (4)

Top-Priority PRs

PR	Title	Score	Action
#51446	Add explicit end marker syntax for inline skills and sub-agents	78	fast_track
#51465	Store sandbox.agent.runtime in aw_info.json and add --runtime filter	62	batch_review
#51472	Deduplicate commit-ref SHA resolution in action_resolver	47	defer (draft)
#51473	Unify skill reference parsing and classification	45	defer (draft)
#51469	Add anti-probing guidance to submit_pull_request_review	30	defer (draft)
#51484	[WIP] Fix safe-output run summary misclassifications	15	defer (WIP, no diff)

Auto-Merge Candidates

None. No PR met the low-risk + high-quality + CI-passing bar simultaneously — the two ready (non-draft) PRs both have mergeable_state issues (blocked/unstable) or unresolved CI status.

Fast-Track Items

• Add explicit end marker syntax for inline skills and sub-agents #51446 — Explicit end-marker syntax for inline skills/sub-agents. All CI checks green (build, lint, unit/JS tests, security scan), has an ADR, well-tested new capability. Recommend expedited human review given parser + generated-workflow-file surface area.

Batch Opportunities

• Store sandbox.agent.runtime in aw_info.json and add --runtime filter to logs/audit #51465 is a strong candidate to batch with any other pending .lock.yml-regeneration PRs (it alone touches 300+ generated lock files from a small compiler change). No other open PRs currently match this pattern, so it stands alone for now but is tagged for future batching.
• The two draft refactor PRs (Deduplicate commit-ref SHA resolution in action_resolver #51472, Unify skill reference parsing and classification #51473) target overlapping files in pkg/workflow/ (skill/action resolution) — worth reviewing together once both leave draft, but not yet batched since neither is ready.

Close Candidates

None — all 6 PRs are active, non-stale, and not superseded.

Full per-PR detail

#51484 — [WIP] Fix safe-output run summary misclassifications

• category: chore (WIP) · risk: low · score: 15 · action: defer
• 0 files changed yet, draft, copilot check in_progress. No diff to review; re-triage once populated.

#51473 — Unify skill reference parsing and classification

• category: refactor · risk: medium · score: 45 · action: defer
• 4 files, +90/-44, draft. Introduces shared parseSkillRefSpec classifier consumed by validation, resolution, and checkout logic. No CI data yet (draft).

#51472 — Deduplicate commit-ref SHA resolution in action_resolver

• category: refactor · risk: medium · score: 47 · action: defer
• 3 files, +84/-16, draft. Consolidates two drifted SHA-resolution code paths into resolveCommitRefSHA; author explicitly flags behavior deltas (Output() vs CombinedOutput(), forced GH_HOST). No CI data yet (draft).

#51469 — Add anti-probing guidance to submit_pull_request_review tool description

• category: docs/chore · risk: low · score: 30 · action: defer
• 2 files, +2/-2, draft. Trivial, safe text addition to both tool-description JSON copies. No CI data yet (draft).

#51465 — Store sandbox.agent.runtime in aw_info.json and add --runtime filter

• category: feature · risk: medium · score: 62 · action: batch_review
• 309 files (300+ regenerated .lock.yml), +724/-13, ready (non-draft). mergeable_state=blocked, commit status pending, no check-run data returned. Adds AgentRuntime field, --runtime filter for logs/audit.

#51446 — Add explicit end marker syntax for inline skills and sub-agents

• category: feature · risk: medium · score: 78 · action: fast_track
• 104 files (~90 regenerated .lock.yml), +1428/-159, ready (non-draft). All core CI green (build/lint/unit/JS tests/security). Includes ADR, new parser logic (pkg/parser/inline_section_helpers.go), JS extractor updates, docs. mergeable_state=blocked — check for conflicts before merge.

Key Trends & Next Actions

1. 4 of 6 PRs are still drafts — CI data unavailable for all of them; re-run triage once they're marked ready.
2. Lockfile noise dominates diff size for both ready PRs (Store sandbox.agent.runtime in aw_info.json and add --runtime filter to logs/audit #51465, Add explicit end marker syntax for inline skills and sub-agents #51446) — actual reviewable surface is much smaller than changed_files suggests; reviewers should focus on the non-.lock.yml files listed above.
3. No OpenTelemetry-related PRs in this batch — the otel/otlp assignment rule did not apply.
4. CI/check-run visibility gap: get_check_runs returned no data for 51465/51472/51473/51469 despite commits existing; only 51446 and 51484 had usable check-run data. Recommend investigating why check-runs aren't populated for some commits (possibly still queued or reported via a different check suite).

Generated by 🔧 PR Triage Agent · auto · 51 AIC · ⌖ 2.57 AIC · ⊞ 8K · ◷

• expires on Aug 9, 2026, 4:40 PM UTC-08:00

[github-actions]

This issue is being closed as outdated. A newer issue has been created: #51544

View newer issue

---

This action was performed automatically by the PR Triage Agent workflow.