Joan León Tag: ai
All the articles with the tag "ai".
A skill that knows nothing: the architecture of Modern Web Guidance
Published:Modern Web Guidance doesn't keep the knowledge in the skill, it teaches the agent to retrieve it. Local semantic search with MiniLM and TensorFlow.js, guides with their own grader, and empirical calibration of the content.
Antigravity Skills: advanced YAML frontmatter configuration
Published:The kind, model, temperature, max_turns and tools fields in SKILL.md frontmatter let you precisely control how each agent runs in Antigravity. A practical guide to the extended fields that Google hasn't officially documented yet.
AI in Web Performance Audits: Deterministic Measurement and Actionable Reports
Published:How to combine deterministic scripts and AI agents to go from metrics collection to an actionable report, and lay the foundations for continuous regression testing in performance.
AI-Driven Performance Debugging: Gemini + Chrome DevTools MCP
Published:A hands-on workshop for performance debugging with Gemini CLI connected to the browser via Chrome DevTools MCP: from identifying a Long Task in minified code to having the agent create the branch and open the PR.
Adding an AI summary to blog posts with the Chrome Summarizer API
Published:How to add an AI summary button to an Astro blog using the browser's native Chrome Summarizer API, without servers or external APIs.
WebPerf Snippets: progressive disclosure in Claude Skills
Published:Applying Anthropic's progressive disclosure system (L1/L2/L3) to reduce the context loaded by WebPerf Snippets Skills and free space in the agent's context window.
WebPerf Snippets agent-first: structured returns and minimal tokens
Published:WebPerf Snippets scripts now return structured JSON directly from evaluate_script() and are minified at build time with dead code elimination, significantly reducing token footprint.
WebPerf Snippets and Agent SKILLs: deterministic audits with Chrome DevTools MCP
Published:WebPerf Snippets become Agent SKILLs with scripts as deterministic tools for AI agents, executable directly in the browser via Chrome DevTools MCP.