nemoclaw-maintainer-verify-stale

Verifies whether stale NemoClaw bug reports still reproduce on the latest release tag.

Updated Aug 1, 2026
One-click install
npx skills add https://github.com/Army161/NemoClaw --skill nemoclaw-maintainer-verify-stale-army161
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: nemoclaw-maintainer-verify-stale
Source: https://github.com/Army161/NemoClaw/tree/main/.agents/skills/nemoclaw-maintainer-verify-stale
Command: npx skills add https://github.com/Army161/NemoClaw --skill nemoclaw-maintainer-verify-stale-army161

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Maintainers of the NVIDIA/NemoClaw repository accumulate stale bug reports that may already be fixed, and manually re-testing each old issue against the latest release is slow, costly, and inconsistent. This Skill automates the verification loop: it selects eligible bug issues, reproduces them on the reported version and the latest tag, and drafts an evidence-backed verdict comment for maintainer approval. ## Core Features & Use Cases - Candidate selection and filtering: Finds open issues with native Issue Type Bug, applies platform, security, integration, idempotency, and active-discussion filters, and parses the reported version from labels, body, or comments. - Baseline-plus-latest reproduction: Extracts or synthesizes a reproducer, validates it against the reported version first, then re-runs it on the latest tag using local-first execution or approved Brev cloud boxes, with special rubrics for performance, rebuild-cycle, log-only, and by-design bugs. - Governed write-back: Scores confidence, redacts secrets, drafts concise verdict comments with tag-anchored links, and applies Project field updates only after explicit maintainer approval; it never auto-closes issues. - Use Case: A maintainer runs a weekly batch to drain the bug backlog, and the Skill verifies up to 15 stale issues, posting redacted verdict comments such as fixed-on-latest, still-reproduces, by-design, or verify-inconclusive. ## Quick Start Ask the agent to verify whether stale NemoClaw bug issue number 2007 still reproduces on the latest release tag.

Frequently Asked Questions about nemoclaw-maintainer-verify-stale

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

FAQPage Schema
How do I verify if a stale GitHub bug report still reproduces?

Provide the issue number and the Skill fetches the issue, parses the reported NemoClaw version, extracts or synthesizes a reproducer, validates it on the reported version, then re-runs it on the latest tag. It drafts a verdict comment with evidence for your approval before posting anything.

How does the Skill reproduce bugs without local hardware?

It provisions Brev cloud instances, preferring to reuse warm verify-stale boxes and capping concurrency at four. Pure CLI bugs are tried locally first for free, and Brev is only used after the maintainer approves the plan and its cost.

Which NemoClaw issues are eligible for stale verification?

The issue must have native Issue Type Bug, not be in Done, Won't Fix, or Duplicate Project status, and lack security, unsupported-platform, integration, or TUI labels. It must also have a parseable reported version that is at least two versions behind latest or older than seven days.

Does the Skill close issues or post comments automatically?

No. It never auto-closes issues and never writes Project fields or public comments before the maintainer explicitly approves the proposed write set. Verdicts are recorded in comments and durable hidden markers, not labels.

What are the limitations of Brev-based bug verification?

Windows, WSL, macOS, and Jetson platform issues are skipped because Brev has no equivalent hardware, and Slack, Discord, Telegram, and similar integration bugs are deferred. TUI or interactive-UI bugs without a non-interactive reproducer are also out of scope in v1.

How are API keys handled during provider-specific reproduction?

Keys are never placed on command lines; they are read via a hidden prompt into a 600-permission file and copied to the Brev box over encrypted SSH. If no key is provided, the Skill substitutes Ollama with a confidence penalty or marks the issue verify-inconclusive.