agent capabilities
9 articles · 11 co-occurring · 0 contradictions · 5 briefs
Today I handed Claude a document that I've been growing for...years on building an orchestrator/distributed runtime that I had only purely theorized possible. One we've been working towards. It would
Today I handed Claude a document that I've been growing for...years on building an orchestrator/distributed runtime that I had only purely theorized possible. One we've been working towards. It would
Agent 需验证计划再执行，降低破坏风险 -- 直接执行 → --dry-run 预览变更" — Introduces dry-run pattern as critical capability for agents to plan before destructive operations. This extends understanding of how agents mitigate
These results mean that there is a hidden mathematical capability in smaller models. But it's gated by the lack of proper representation!" — Article provides concrete empirical evidence (47%→99% on GP
Coding agents suck at using a browser" — Article identifies a specific capability gap in agents (browser interaction) and presents Dev Browser skill as a solution, demonstrating how skills extend agen
extension to share current VSCode open files and selected lines with pi coding agent" — Demonstrates extending agent capabilities by providing access to IDE context (open files and selections)
Give the skill to your openclaw buddies and let them cook" — The extension extends agent capabilities by providing access to summarization, content drafting, and social media generation skills
We built a benchmark to find out" — Article provides empirical benchmark methodology for measuring agent capability on real-world integration tasks
[INFERRED] "Openclaw controling my Blender" — Demonstrates agent capability extension: OpenClaw gains design/3D manipulation ability through Blender integration, expanding beyond pure software tasks
[INFERRED] "prompts -> rules -> skills" — Article describes a progression model for building agent capabilities from prompts through rules to skills, suggesting a layered capability development strate