Team Adoption of Claude Skills: Habits That Stick
The habits, norms, and change management that make Claude Agent Skills stick across a team — and the pitfalls that leave good skills unused.
Agentic AI, LLM engineering, and the models behind modern automation — multi-agent systems, LLM evaluation and comparisons, RAG, fine-tuning, AI infrastructure, security, and production AI engineering.
From the blog
The habits, norms, and change management that make Claude Agent Skills stick across a team — and the pitfalls that leave good skills unused.
Where time and money savings from Claude Agent Skills actually come from — a grounded cost model, what to measure, and the pitfalls that erase ROI.
A staged playbook to migrate an existing workflow onto Claude agents — strangler-fig rollout, shadow mode, human-in-the-loop, and reversible cutover.
Build an eval loop for Claude agents — task-level metrics, LLM-as-judge, regression suites, and release gates that stop bad changes from shipping.
Security hardening for Claude agents — sandbox execution, least-privilege tools, secret protection, and prompt-injection defense for tool-using systems.
Lower Claude agent cost and latency with prompt caching, batching, context pruning, and model routing — concrete tactics and honest tradeoffs.
Diagnose and fix the three big Claude agent failures — infinite loops, wrong tool calls, and hallucinated arguments — with traces, schemas, and guardrails.
Prompt and context design for Claude agents — the three tiers of context, what to include, what to compute with tools, and what to leave out, and why.
Connect MCP servers to Claude Agent Skills with sound auth, schema design, structured error handling, and idempotency so your agents act reliably in production.