nm-conserve-cpu-gpu-performanceEstablishes CPU/GPU baselines before resource-intensive operations. Use before builds, training runs, or any task that pins cores or GPUs for over a minute
Install via ClawdBot CLI:
clawdbot install athola/nm-conserve-cpu-gpu-performanceGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Calls external URL not in known-safe list
https://github.com/athola/claude-night-market/tree/master/plugins/conserveAudited Apr 16, 2026 · audit v1.0
Generated Oct 8, 2026
An internal ML platform team auto-loads this skill before every training job submission to capture GPU baselines, narrow smoke tests, and enforce per-job GPU-minute budgets. Profiling with nvidia-smi dmon reveals idle time and memory pressure before scaling to full sweeps. This prevents runaway CI bills and node starvation across shared clusters.
A game studio uses the skill before nightly C++ builds and asset pipeline runs that pin all CPU cores. Engineers run perf and flamegraph to prove hot paths, then throttle via nice/ionice and diff-based bazel targets. The result is shorter build queues and fewer blocked workstations.
A university HPC lab adopts the skill as a pre-flight checklist for Slurm jobs, recording GPU-hour quotas and choosing unit-scale inputs before full training runs. Profilers like nsys and DLProf capture kernel timelines, guiding staged escalation from failed tests to full sweeps. Researchers stay within grant budgets while avoiding starvation of shared nodes.
A risk modeling team runs this skill before retraining fraud detection models on GPU clusters, baselining utilization and setting per-task GPU-minute budgets. They use selective pytest -k validation and checkpoint/restore instead of full restarts. This keeps training windows within compliance quotas and reduces wasted compute spend.
A platform engineering team embeds the skill into shared CI templates so every heavy job starts with a baseline, scope narrowing, and instrumented run. They log decisions and resource costs per PR, making CPU/GPU budget overruns visible. This evolves into a culture of evidence-based throttling and staged test escalation.
The skill is distributed freely as part of the open-source Night Market plugin, while enterprise features like centralized budget dashboards, team-level audit logs, and policy enforcement are sold as a paid add-on. Companies pay for compliance and cost visibility across many teams.
Offer expert engagements that embed cpu-gpu-performance practices into existing CI/CD, Slurm/Kubernetes, and training pipelines, including custom profiler instrumentation and budget policy design. Clients pay for faster onboarding and measurable compute savings.
A hosted service that automatically applies baseline capture, scope narrowing, and throttling rules to customer jobs, charging based on compute volume or number of managed nodes. It bundles profiling dashboards and budget alerts as a SaaS product.
💬 Integration Tip
Prepend the skill to CI job templates or a SessionStart hook so baseline and budget checks run automatically before any heavy command. Start by logging resource cost for one job, then gradually add profiler instrumentation and throttling rules as teams see the savings.
Scored Oct 8, 2026
Firecrawl CLI for web scraping, crawling, and search. Scrape single pages or entire websites, map site URLs, and search the web with full content extraction. Returns clean markdown optimized for LLM context. Use for research, documentation extraction, competitive intelligence, and content monitoring.
Incident Commander Skill
Monitors OpenClaw cron job health, identifies failures, timeouts, and delivery issues.
Observability Designer (POWERFUL)
Use when defining SLIs/SLOs, managing error budgets, or building reliable systems at scale. Invoke for incident management, chaos engineering, toil reduction, capacity planning.
QQ bot integration for OpenClaw with session isolation, logging, and AI auto-reply. Supports private chat and group @messages.