Manages GPU profiling workflows and performance evidence capture via Nsight.
Install
mkdir -p .claude/skills/mission-control-nsight-profiling && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12915" && unzip -o skill.zip -d .claude/skills/mission-control-nsight-profiling && rm skill.zipInstalls to .claude/skills/mission-control-nsight-profiling
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Coordinate Nsight-based GPU profiling through Mission Control with safe evidence capture and approval-aware command use.Key capabilities
- →Read Mission Control status, validation recipe, and diagnostics.
- →Request Mission Control to isolate GPU commands or tests for profiling.
- →Capture profiler results as evidence.
- →Feed profiling results into optimization or review steps.
- →Summarize profiling results in bridge-safe language.
How it works
This skill coordinates Nsight-based GPU profiling through Mission Control, ensuring profiling commands adhere to approval policies and results are captured as evidence. It translates profiler output into actionable steps.
Inputs & outputs
When to use mission-control-nsight-profiling
- →Profiling gpu performance
- →Capturing performance evidence
- →Analyzing cuda bottlenecks
About this skill
Mission Control Nsight Profiling
Purpose
Use Mission Control to plan or run Nsight profiling for GPU code paths and turn profiler output into actionable next steps.
The Codex chat agent is not the Mission Control Manager. It is the bridge between the user and the Mission Control Manager.
Use when
- The user asks for GPU profiling.
- A CUDA change needs Nsight Systems or Nsight Compute evidence.
- Performance claims need more than benchmark anecdotes.
Workflow
- Read the current Mission Control status, validation recipe, and diagnostics.
- Ask Mission Control to isolate the GPU command or test worth profiling.
- Keep profiler commands behind approval policy when needed.
- Capture the profiler result as evidence, not raw noise.
- Feed the result back into the next optimization or review step.
Mission Control calls
Tools:
mission_control_start_taskmission_control_get_statusmission_control_get_handoff_summary
Resources:
mission-control://projects/{project_id}/validation-summarymission-control://projects/{project_id}/diagnosticsmission-control://projects/{project_id}/handoff
User-facing output
- Show the target GPU path Mission Control wants to profile.
- Show the profiler loop and next expected evidence.
- Summarize the profiling result in bridge-safe language.
Approval behavior
Profiler commands and any environment changes still follow Mission Control approval policy.
Never do
- Do not dump raw profiler spam into chat.
- Do not claim a bottleneck moved unless the profile actually shows it.
- Do not profile the wrong workload just because it is convenient.
Failure and fallback
If Nsight is unavailable, report that honestly and fall back to benchmark comparison plus code-path review.
Example invocation
Use Mission Control to profile the current CUDA path with Nsight and summarize the bottleneck.
When not to use it
- →When dumping raw profiler spam into chat.
- →When claiming bottleneck movement without profile evidence.
- →When profiling an incorrect workload.
Limitations
- →Profiler commands follow Mission Control approval policy.
- →Environment changes follow Mission Control approval policy.
- →If Nsight is unavailable, it reports that and falls back to benchmark comparison plus code-path review.
How it compares
This skill integrates Nsight profiling with Mission Control, providing a structured, approval-aware workflow for GPU performance analysis and evidence capture, unlike direct profiler execution.
Compared to similar skills
mission-control-nsight-profiling side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| mission-control-nsight-profiling (this skill) | 0 | 2mo | No flags | Advanced |
| chrome-devtools | 41 | 7mo | Review | Intermediate |
| python-performance-optimization | 27 | 2mo | No flags | Intermediate |
| analyzing-logs | 14 | 27d | Review | Beginner |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by MN755
View all by MN755 →You might also like
chrome-devtools
mrgoonie
Browser automation, debugging, and performance analysis using Puppeteer CLI scripts. Use for automating browsers, taking screenshots, analyzing performance, monitoring network traffic, web scraping, form automation, and JavaScript debugging.
python-performance-optimization
wshobson
Profile and optimize Python code using cProfile, memory profilers, and performance best practices. Use when debugging slow Python code, optimizing bottlenecks, or improving application performance.
analyzing-logs
jeremylongshore
Analyze application logs to detect performance issues, identify error patterns, and improve stability by extracting key insights.
redis-inspect
civitai
Inspect Redis cache keys, values, and TTLs for debugging. Supports both main cache and system cache. Use for debugging cache issues, checking cached values, and monitoring cache state. Read-only by default.
obsidian-performance-tuning
jeremylongshore
Optimize Obsidian plugin performance for smooth operation. Use when experiencing lag, memory issues, or slow startup, or when optimizing plugin code for large vaults. Trigger with phrases like "obsidian performance", "obsidian slow", "optimize obsidian plugin", "obsidian memory usage".
turborepo-caching
wshobson
Configure Turborepo for efficient monorepo builds with local and remote caching. Use when setting up Turborepo, optimizing build pipelines, or implementing distributed caching.