[Weekly Summary] Weekly Issue Summary — Oct 12–19, 2026

[github-actions[bot]]

Overview

From 2026-10-12 through 2026-10-19, github/gh-aw opened 1,731 issues. Almost all of that volume is automation: 1,694 issues (97.9%) came from app/github-actions. 37 issues were opened by humans. 1,506 of the week's issues are already closed (87.0% close rate); 225 remain open, including 1 human-authored item.

The 30-day window behind the charts (Sep 19–Oct 19) shows a slight backlog drawdown: 6,035 opened vs 6,082 closed (net -47), with open issues moving from about 314 to 267.

Key metrics

Metric	Value
Issues opened (7d)	1,731
Closed among those	1,506 (87.0%)
Still open among those	225
Human-authored	37 (SivaKesava1 18, lpcox 12, pelikhan 6, davidslater 1)
Bot-authored	1,694 (app/github-actions)
Median time-to-close (week's closed issues)	~0.14 days
Mean time-to-close (week's closed issues)	~0.38 days
Open issues now (repo-wide)	267
30-day net (opened − closed)	-47

Opened per day (7d)

Date	Opened	Human
2026-10-12	251	11
2026-10-13	262	4
2026-10-14	199	8
2026-10-15	188	4
2026-10-16	246	5
2026-10-17	231	4
2026-10-18	178	1
2026-10-19 (partial UTC)	176	0

📈 Issue Activity Trends

Weekly Activity Patterns

Throughput is high and fairly stable at ~1,200–1,550 issues/week. Closures have outpaced openings since the week of Sep 28, which pulled the open backlog down from a peak near 351 to the mid-230s before a partial rebound on Oct 19 (still in progress). Net change is small relative to volume, so the backlog is being held, not exploding.

Resolution Time Analysis

Most issues close the same day they open (typical of agentic/automation traffic). Daily average time-to-close stays well under 1 day; the 7-day moving average is only ~4% higher than the prior week. The shaded IQR stays tight, with occasional spikes when longer-lived items land in a day's close set. Human bugs and routing work still close quickly in this sample.

Notable patterns

• Automation dominates. Top title prefixes: [aw] (542), [WIP] (457), unprefixed (309), [squad:plan] (67), [PR Triage Report] (28), [work queue] (24), [avenger] (19).
• Top labels (7d): agentic-workflows (583), automation (428), testing (263), cascade-suspected (212), cookie (126), squad:plan (67), security (43).
• Human themes: model-routing (18 of 37 human issues), firewall/mcpg version bumps, NVX/Copilot guest runtime, safe-output / transport-integrity, and one P0 provenance bug (#68075).
• Still open from this week: 225 items, almost all bot workflows. Only human leftover is #69180 (AWF redact_secrets step). Recurring open bot clusters: [aw] (39), [work queue] (24), [avenger] (13), [WIP] / [AW Top 10] (10 each).
• cascade-suspected (212) is a large slice of bot traffic — worth watching so suspected cascades do not hide real product defects.

Human-authored issues (full list)

37 issues from SivaKesava1, lpcox, pelikhan, and davidslater. All but #69180 are already closed.

37 human-authored issues opened Oct 12–19

• NVX guest command requires unavailable RUNNER_TOOL_CACHE and RUNNER_TEMP assets #67824 CLOSED @lpcox — NVX guest command requires unavailable RUNNER_TOOL_CACHE and RUNNER_TEMP assets (bug, compiler)
• NVX must select Copilot engine assets compatible with guest libc #67892 CLOSED @lpcox — NVX must select Copilot engine assets compatible with guest libc (bug, copilot, compiler)
• Routed Claude and Codex workflows compile without model-provider: github, then fail for a missing provider key #68066 CLOSED @SivaKesava1 — Routed Claude and Codex workflows compile without model-provider: github, then fail for a missing provider key (model-routing)
• Routed smokes: cover Claude main models on the Copilot and pi engines, the SDK reverse direction, and non-economy modes #68067 CLOSED @SivaKesava1 — Routed smokes: cover Claude main models on Copilot and pi, SDK reverse direction, and non-economy modes (model-routing)
• Routed smokes: cover Claude main models on the Copilot and pi engines, the SDK reverse direction, and non-economy modes #68068 CLOSED @SivaKesava1 — Routed smokes: cover Claude main models on Copilot and pi, SDK reverse direction, and non-economy modes (model-routing)
• Run the routed smokes on pull requests and merges that touch routing code #68069 CLOSED @SivaKesava1 — Run the routed smokes on pull requests and merges that touch routing code (model-routing)
• Run the routed smokes on pull requests and merges that touch routing code #68074 CLOSED @SivaKesava1 — Run the routed smokes on pull requests and merges that touch routing code (model-routing)
• Always preserve execution provenance in safe outputs, including footer:false #68075 CLOSED @davidslater — Always preserve execution provenance in safe outputs, including footer:false (bug, security, safe-outputs, P0)
• Add golden fixtures for routed failures and the routed paths that have none #68076 CLOSED @SivaKesava1 — Add golden fixtures for routed failures and the routed paths that have none (model-routing)
• Improve diagnostics when daily AIC accounting artifacts cannot be read #68078 CLOSED @lpcox — Improve diagnostics when daily AIC accounting artifacts cannot be read
• Pi engine fails at startup: pi-coding-agent@1.1.0 resolves pi-codemode@1.2.0, which removed MCP_TYPESCRIPT_PREAMBLE #68172 CLOSED @SivaKesava1 — Pi engine fails at startup: pi-coding-agent@1.1.0 resolves pi-codemode@1.2.0, which removed MCP_TYPESCRIPT_PREAMBLE (model-routing)
• Run the Copilot, SDK and Claude routed smokes on their 2-day schedule only, like smoke-pi-routed #68264 CLOSED @SivaKesava1 — Run Copilot, SDK and Claude routed smokes on their 2-day schedule only (model-routing)
• Bump gh-aw-firewall to v0.28.56 #68335 CLOSED @lpcox — Bump gh-aw-firewall to v0.28.56
• Model routing: make models.allowed / models.blocked the only request policy, and allowed-models optional router candidates #68494 CLOSED @SivaKesava1 — Model routing: make models.allowed / models.blocked the only request policy (model-routing)
• Model routing: an AI-credit budget failure after a routed selection is recorded as harness_selection_mismatch #68497 CLOSED @SivaKesava1 — AI-credit budget failure after routed selection recorded as harness_selection_mismatch (model-routing)
• Bump gh-aw-mcpg to v0.4.31 and gh-aw-firewall to v0.28.57 #68597 CLOSED @lpcox — Bump gh-aw-mcpg to v0.4.31 and gh-aw-firewall to v0.28.57
• Safe-output integrity checks block indefinitely when a captured artifact becomes a FIFO #68786 CLOSED @pelikhan — Safe-output integrity checks block indefinitely when a captured artifact becomes a FIFO
• Permanent transport-integrity errors can enter API rate-limit retries based on branch names #68787 CLOSED @pelikhan — Permanent transport-integrity errors can enter API rate-limit retries based on branch names
• Transport-integrity failures prevent local Git cleanup and leave temporary bundle refs #68788 CLOSED @pelikhan — Transport-integrity failures prevent local Git cleanup and leave temporary bundle refs
• Safe-output integrity guards repeatedly read and hash full bundles for unrelated Git commands #68789 CLOSED @pelikhan — Safe-output integrity guards repeatedly read and hash full bundles for unrelated Git commands
• Transport-integrity errors discard filesystem causes needed for debugging #68790 CLOSED @pelikhan — Transport-integrity errors discard filesystem causes needed for debugging
• Claude on Copilot: pass Claude Code a modelOverrides map for Copilot's dotted Claude model ids (smoke-claude-on-copilot 400 invalid_reasoning_effort) #68834 CLOSED @SivaKesava1 — Claude on Copilot: pass Claude Code a modelOverrides map for dotted Claude model ids (claude, model-inventory, model-routing)
• Safe-outputs MCP gateway fails when GH_AW_TARGET_REPO_SLUG is unset #68881 CLOSED @lpcox — Safe-outputs MCP gateway fails when GH_AW_TARGET_REPO_SLUG is unset
• Dry-run scanner gate fails on compiler-generated empty env and removed daily-credit output references (v0.92.0) #69078 CLOSED @lpcox — Dry-run scanner gate fails on compiler-generated empty env and removed daily-credit output references (v0.92.0)
• audit/logs: sub-agent cost rows lose served models for declared-model ids and are keyed by display name instead of agent definition #69080 CLOSED @SivaKesava1 — audit/logs: sub-agent cost rows lose served models for declared-model ids (model-routing)
• Safely redact and collect AWF's early AWF_RESOLVED_CONFIG artifact after failed cleanup #69107 CLOSED @lpcox — Safely redact and collect AWF early AWF_RESOLVED_CONFIG artifact after failed cleanup
• Bump gh-aw-firewall to v0.28.59 and gh-aw-mcpg to v0.4.33 #69108 CLOSED @lpcox — Bump gh-aw-firewall to v0.28.59 and gh-aw-mcpg to v0.4.33
• Copilot workflow compatibility regressions with pinned CLI versions and Cloud Hypervisor metadata writes #69161 CLOSED @lpcox — Copilot workflow compatibility regressions with pinned CLI versions and Cloud Hypervisor metadata writes (bug)
• AWF artifact upload guards reference missing redact_secrets step #69180 OPEN @pelikhan — AWF artifact upload guards reference missing redact_secrets step
• pi: bare mode hides workflow inline skills and can fail Claude-main runs #69427 CLOSED @SivaKesava1 — pi: bare mode hides workflow inline skills and can fail Claude-main runs (model-routing)
• audit/logs: reconcile legacy sub-agent lifecycle and metrics when unified events lack agentId #69430 CLOSED @SivaKesava1 — audit/logs: reconcile legacy sub-agent lifecycle and metrics when unified events lack agentId (model-routing)
• Resolve model: aliases the same way on every engine #69484 CLOSED @SivaKesava1 — Resolve model aliases the same way on every engine (model-routing)
• Bump gh-aw-firewall to v0.28.62 #69525 CLOSED @lpcox — Bump gh-aw-firewall to v0.28.62
• Bump MCP Gateway (mcpg) to v0.4.35 #69580 CLOSED @lpcox — Bump MCP Gateway (mcpg) to v0.4.35
• Preserve AWF candidate rankings and request/failure metadata in unified sessions (G-2/G-11) #69725 CLOSED @SivaKesava1 — Preserve AWF candidate rankings and request/failure metadata in unified sessions (model-routing)
• Preserve AWF model inventories with null endpoints and live reasoning metadata (G-7) #69754 CLOSED @SivaKesava1 — Preserve AWF model inventories with null endpoints and live reasoning metadata (model-routing)
• Copilot CLI model-selection persistence aborts routed attribution and token usage #69965 CLOSED @SivaKesava1 — Copilot CLI model-selection persistence aborts routed attribution and token usage (bug, model-routing)

Bot-authored issues

1,694 issues from app/github-actions. Listing every title would be unreadable; grouped by prefix instead.

Bot title prefixes (7d)

Prefix	Count
[aw]	542
[WIP]	457
(no prefix)	309
[squad:plan]	67
[PR Triage Report]	28
[work queue]	24
[avenger]	19
[pr-sous-chef]	14
[AW Top 10]	12
[lint-monster]	11
[container-image-scan]	11
[aw-compat]	9
[static-analysis]	9
[team-status]	8
[work-queues]	8
[ca]	8
[github-docs-seo]	8
[copilot-cli-research]	8
[Contribution Check Report]	8
[smoke-temporary-id]	8
Other prefixes	remainder

Next actions

• Close or triage AWF artifact upload guards reference missing redact_secrets step #69180 (only human issue still open from this week).
• Review remaining [work queue] / [avenger] / [AW Top 10] opens so agent queues do not accumulate.
• Spot-check cascade-suspected (212) for real regressions vs noisy follow-on issues.
• Human work is concentrated on model-routing and firewall/mcpg bumps — keep those smoke/routing fixtures green.

References:

• §39273204856

Generated by 📅 Weekly Issue Summary · pi · gro46 · 39.7 AIC · ⌖ 17.1 AIC · ⊞ 10.5K · ◷

• expires on Oct 20, 2026, 7:07 AM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-10-20T15:07:52.410Z.

Closed by Workflow