[deep-report] Fix reviewer-squad gate misreporting bot-PR skips as failures (30/121 daily failures)

[github-actions]

Description

The Agentic Workflow Audit Agent (discussion #56459, 2026-08-27/28) found that 6 PR-review workflows — Design Decision Gate 🏗️, Impeccable Skills Reviewer, Matt Pocock Skills Reviewer, PR Code Quality Reviewer, Ponytail Reviewer, Test Quality Sentinel — fail in ~14 seconds on every PR opened by the GitHub Copilot coding agent (branch prefixed copilot/*, actor=null, no engine ever invoked). This looks like a skip/gate condition exiting non-zero instead of reporting neutral/skipped, and accounted for 30 of the day's 121 raw fleet failures (24.8%).

Expected Impact

Fixing the gate condition to report neutral/skipped (instead of failure) for intentionally-skipped bot-authored PRs removes 30/121 raw failures fleet-wide, moving the adjusted daily success rate from 70.28% → 75.5% with a single shared-condition fix. Also stops noisy false-positive failure notifications for maintainers watching these workflows.

Suggested Fix

Locate the shared gate/skip condition used by these 6 review workflows (likely a common step or reusable check for bot-authored PRs) and change its non-matching-condition branch to exit with a neutral/skipped conclusion rather than a failure exit code. Verify via a sample Copilot-authored PR run that the job now reports skipped instead of failure.

Suggested Agent

New Agent (workflow YAML/config fix, touches a shared gate pattern across 6 lock files)

Estimated Effort

Quick (< 1 hour)

Data Source

DeepReport analysis of Agentic Workflow Audit Agent (discussion #56459, 2026-08-27/28); example run https://github.com/github/gh-aw/actions/runs/33127429505

Generated by 🔬 Deep Report · claude · agent · 103.9 AIC · ⌖ 9.24 AIC · ⊞ 12.4K · ◷

• expires on Aug 29, 2026, 7:32 PM UTC-08:00

[github-actions]

Corroborating data — cycle 2026-08-29

This cycle's fleet-health pass (Audit Workflows discussion #56739) again shows a large bloc of PR-review workflows failing together in 1-9 seconds with no engine log populated — this run: 57/93 daily failures attributed to the same signature (previous cycle: 30/121).

Root cause, now verified by direct grep (not just inferred from failure timing): all of the affected workflows —
design-decision-gate.md, impeccable-skills-reviewer.md, pr-code-quality-reviewer.md, ponytail-reviewer.md, test-quality-sentinel.md — import shared/pr-review-base.md, which sets:

import-schema:
  min-integrity:
    type: string
    default: "approved"
imports:
  - shared/github-guard-policy.md

and shared/github-guard-policy.md sets:

tools:
  github:
    approval-labels: [cookie, community]

So any PR lacking the cookie/community label (or sufficient author association) trips the GitHub MCP min-integrity: approved gate before the agent engine ever starts, producing an instant, no-log, all-or-nothing failure that the fleet dashboards currently count as an agent failure rather than a policy-gate skip.

Filing this as a comment rather than a new issue since this is the same pattern already tracked here — just adding today's numbers and the confirmed shared-import root cause to help prioritize a fix (e.g. excluding gate-skips from the failure-rate metric, or auto-labeling qualifying PRs earlier in the pipeline).

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• api.anthropic.com

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "api.anthropic.com"

See Network Configuration for more information.

Generated by 🔬 Deep Report · claude · agent · 258.7 AIC · ⌖ 8.26 AIC · ⊞ 12.4K · ◷

[github-actions]

This issue was automatically closed because it expired on 2026-08-30T03:32:32.715Z.

Closed by Workflow