Safe Debug
Skill Verified ActiveTrusted-lane debug skill for deep learning research work. Use when the user pastes a traceback, terminal error, CUDA OOM, checkpoint load failure, shape mismatch, NaN loss symptom, or training failure and wants conservative diagnosis before any patching. Do not use for broad refactoring, speculative adaptation, automatic exploratory patching, or general repository familiarization.
To provide a safe and conservative debugging experience for deep learning research, ensuring that diagnoses are thorough and any proposed code changes are minimal and explicitly approved.
Features
- Conservative error diagnosis for DL research failures
- Classifies errors into categories (CUDA OOM, checkpoint mismatch, etc.)
- Suggests minimal, safe fixes and recovery steps
- Requires explicit user approval before patching code
- Outputs detailed diagnosis and patch plan
Use Cases
- Diagnosing CUDA Out Of Memory errors
- Analyzing checkpoint loading failures
- Troubleshooting shape mismatches in model tensors
- Investigating NaN loss symptoms during training
- Understanding terminal errors and tracebacks
Non-Goals
- Performing broad repository refactoring
- Automatic exploratory patching
- General repository familiarization without a failure symptom
- Speculative adaptation of code
Installation
npx skills add lllllllama/ai-paper-reproduction-skillRuns the Vercel skills CLI (skills.sh) via npx — needs Node.js locally and at least one installed skills-compatible agent (Claude Code, Cursor, Codex, …). Assumes the repo follows the agentskills.io format.
Quality Score
VerifiedSimilar Extensions
Trajectory Review
99Post-hoc diagnosis of a failed agent trajectory. Classifies the first unrecoverable step into one of nine failure categories (plan adherence, hallucinated information, invalid tool call, misread tool output, intent–plan mismatch, under-specified intent, unsupported intent, guardrail trigger, system failure) and produces an evidence-backed root-cause report.
Node Connect
100Diagnose OpenClaw Android, iOS, or macOS node pairing, QR/setup code, route, auth, and connection failures.
Openclaw Debugging
100Debug OpenClaw model, provider, tool-surface, code-mode, streaming, and live/Crabbox behavior by choosing the right logs, probes, and proof path before changing code.
Systematic Debugging
100Systematic debugging methodology emphasizing root cause analysis over quick fixes
Troubleshooting
100Uses Chrome DevTools MCP and documentation to troubleshoot connection and target issues. Trigger this skill when list_pages, new_page, or navigate_page fail, or when the server initialization fails.
Root Cause Tracing
99Use when errors occur deep in execution and you need to trace back to find the original trigger - systematically traces bugs backward through call stack, adding instrumentation when needed, to identify source of invalid data or incorrect behavior