nemoclaw-maintainer-fix-e2e-failures

Coordinates continuous root-cause fixes for automatic NemoClaw E2E failures on main.

22.3k|3.1k|Updated Mar 15, 2026
One-click install
npx skills add https://github.com/NVIDIA/NemoClaw --skill nemoclaw-maintainer-fix-e2e-failures
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: nemoclaw-maintainer-fix-e2e-failures
Source: https://github.com/NVIDIA/NemoClaw/tree/main/.agents/skills/nemoclaw-maintainer-fix-e2e-failures
Command: npx skills add https://github.com/NVIDIA/NemoClaw --skill nemoclaw-maintainer-fix-e2e-failures

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Automatic end-to-end test runs on the NemoClaw main branch fail for many overlapping reasons, and uncoordinated fixes lead to duplicate PRs, stale approvals, and unsafe merges. This Skill runs a continuous maintainer loop that groups failures by root cause, assigns ownership through draft PRs, enforces independent review, and merges only when GitHub rules pass.

Core Features & Use Cases

  • Root-Cause Queue Management: Groups failures by causal signature rather than job name, tracks states from unclaimed to merged, and reconciles claims against open PRs before editing.
  • Policy-Gated GitHub Writes: Runs a trusted policy evaluator before workflow approvals, reviews, merges, retries, and rollback PRs, denying ambiguous or unauthorized writes.
  • Post-Merge Containment: Detects regressions after a merge, blocks related merge writes, and permits only explicitly authorized draft revert PRs.
  • Use Case: A maintainer agent monitors automatic main E2E runs overnight, claims an unowned failure with a draft PR, reviews a peer's fix while waiting for CI, and hands off full queue state to another agent without ending the loop.

Quick Start

Ask the agent to continuously fix failures from automatic NemoClaw E2E runs on main, coordinating with other maintainers and merging eligible PRs.

Frequently Asked Questions about nemoclaw-maintainer-fix-e2e-failures

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I continuously fix E2E failures on the NemoClaw main branch?

Run this maintenance loop to inspect automatic E2E results, group failures by root cause, claim each unowned cause with a draft PR, and merge only after independent approval and required checks pass. The loop has no scheduled end time and keeps monitoring new runs.

How are E2E failures grouped before creating fix PRs?

Failures are grouped by a root-cause key combining affected component, failure phase, and stable causal signature, not by job name. Each group gets exactly one draft PR containing the run ID, failed job IDs, and redacted failure signature.

Can a maintainer approve their own E2E fix PR?

No. The policy evaluator denies self-approval and requires at least one approval of the latest PR commit from an account that did not open, author, or co-author the PR. Review claims are bound to a specific commit SHA.

When should I use nemoclaw-maintainer-e2e instead of this Skill?

Use nemoclaw-maintainer-e2e for manual E2E dispatch, such as running the suite for a release candidate or approving secret-bearing environment jobs. This Skill only observes automatic main runs and never triggers manual duplicates.

What happens when a merged E2E fix causes a regression on main?

The confirming agent becomes containment owner, marks the root cause blocked, and pauses related merge writes. A draft revert PR is allowed only with explicit authorization, and main is never reverted directly.

Does the maintenance loop end when the newest main E2E run passes?

No. A passing run verifies only its tested commit SHA. The loop records the evidence and continues monitoring until the operator cancels it or another agent acknowledges ownership through a continuity handoff.