97 skills · sorted by relevance
Investigate slow TPU training on GKE by comparing diagnostic events with one-minute monitoring metrics. Separate memory or CPU bottlenecks from hardware and network issues, and identify the appropriate next steps.
Diagnose why an AI agent becomes less reliable as conversations and documents accumulate. Distinguish missed information, persistent false claims, irrelevant material, mixed tasks, and conflicting sources, then choose ways to clean, reorganize, or separate the context.
Compare current benchmark results, measurements of performance, with historical baselines to identify degradation between builds. Produce a comparison summary, rank affected metrics by severity, and outline questions for further investigation.
Token-range severity zones (Green/Yellow/Orange/Red/Critical) with detection checklist, early warning indicators, and corrective routing actions for context window degradation
The English summary is being prepared. Showing the original description.
This skill should be used when the user asks to "diagnose context problems", "fix lost-in-middle issues", "debug agent failures", "understand context poisoning", or mentions context degradation, attention patterns, context clash, context confusion, or agent performance degradation. Provides patterns for recognizing and mitigating context failures.
The English summary is being prepared. Showing the original description.
Learn how instructions, documents, tool results, and conversation history form an AI model’s context. Understand attention limits, why information in the middle can be missed, and how relevance and placement inform decisions about what to include.
Handle failed video encoding by checking available encoders, testing a short clip, and trying alternatives. The workflow covers copying without re-encoding, switching codecs, which define video compression, and using MoviePy as another processing option.
Design software that keeps working with reduced features when an optional service is unavailable. The skill provides patterns for early availability checks, cached results, and messages explaining what is missing, which features are affected, and how to restore them.
Graceful Degradation with Helpful Messages
The English summary is being prepared. Showing the original description.
Think and work like an expert Biomaterials Scientist. Use when a task calls for Biomaterials Scientist judgment. Reasons from material-biology interfaces, degradation-product toxicity, and mechanical mismatch to host tissue through ISO 10993 biological evaluation plans, ISO 10993-12 extract conditions, GPC/ICP/SEM characterization, and sterilization validation while treating endotoxin contamination, pH shift from acidic degradation, extract-ratio artifacts, and non-final-form test articles as first-class failure modes.
The English summary is being prepared. Showing the original description.
Work on bifunctional degraders and molecular glues, where potency comes from a ternary complex rather than occupancy. Use this skill to apply the property rules that govern this beyond-rule-of-five space, reason about linker length, attachment vector and E3 ligase choice, prepare inputs for ternary complex structure prediction, and interpret degradation readouts — DC50, Dmax, cooperativity, and the hook effect that makes a dose-response curve turn over. Also trigger on PROTAC, molecular glue, targeted protein degradation, E3 ligase, cereblon, VHL, ternary complex, DC50, Dmax, hook effect, cooperativity, or PROTAC-DB.
The English summary is being prepared. Showing the original description.
Implement comprehensive model drift monitoring using Evidently AI, statistical tests (PSI, KS), and custom metrics to detect data drift and concept drift in production ML systems. Set up automated alerting and reporting workflows to catch degradation before it impacts business metrics. Use when production models show unexplained performance degradation, when new data distributions differ from training data, when seasonal shifts affect input features, or when regulatory requirements mandate model monitoring.
The English summary is being prepared. Showing the original description.
Language models exhibit predictable degradation patterns as context length increases. Understanding these patterns is essential for diagnosing failures and designing resilient systems.
The English summary is being prepared. Showing the original description.
Expert-level battery technology covering electrochemistry, lithium-ion battery physics, battery management systems, degradation mechanisms, and next-generation battery chemistries.
The English summary is being prepared. Showing the original description.
Diagnose AI agent runs that fail repeatedly, loop through tools, or lose track of the task. Capture errors and environment details, test a focused recovery action, and produce a report explaining the evidence and outcome.
Design industrial-style web interfaces using rigid grids, dramatic type sizes, compact data displays, and red accents. Choose a light print-inspired layout or a dark terminal aesthetic, with textures that mimic paper or older screens.
Design application and API behavior when operations fail. Choose between exceptions and Result types, which explicitly represent success or failure, and improve retries, error messages, logging, and resource cleanup so failures are easier to handle and investigate.
Create a structured review after a production incident, focusing on systems and processes without blaming individuals. Reconstruct events, quantify impact, investigate causes, and document specific follow-up actions with named owners and deadlines.
Reduce the memory needed to run large language models with GPTQ quantization, which stores model weights at lower precision. Covers loading compressed models, quantizing your own models, choosing execution backends, and connecting to memory-efficient fine-tuning.
Review board memos and strategy briefs with available Claude, Codex, and Gemini models. Compare concerns, supporting arguments, and votes, then turn disagreements into questions to resolve before committing to a major business decision.
Investigate stalled TPU training across multiple slices on GKE using hang reports, logs, and latency metrics. Separate compilation inconsistencies and data input stalls from hardware or network faults, and identify the relevant nodes for further investigation.
Analyze records of an AI agent’s decisions and tool use to find repeated failures, lost goals, and misunderstood instructions. Use the findings to refine prompts, rerun tasks, and turn debugging lessons into reusable skills.
Compress large language models to 4-bit weights to reduce GPU memory needs and accelerate text generation. Covers loading compressed models, converting your own models, choosing settings, and preparing memory-efficient fine-tuning with transformers and PEFT.
Help an agent periodically review whether its work stays on task, remains thorough, avoids repetition, respects earlier decisions, and covers every item. Record assessments and revisit working notes when quality declines so the approach can be adjusted.