From c862d1208c696f50b1a3d5cc9faf0e08a5b498d3 Mon Sep 17 00:00:00 2001 From: Michele Cereda Date: Fri, 20 Feb 2026 02:19:19 +0100 Subject: [PATCH] chore(kb/ai): review and expand notes --- knowledge base/ai/agent.md | 28 ++++++--- knowledge base/ai/claude/claude code.md | 11 +++- knowledge base/ai/llama.cpp.md | 2 +- knowledge base/ai/llm.md | 18 ++++-- knowledge base/ai/vllm.md | 1 + knowledge base/uv.md | 77 +++++++++++++++++++++++++ 6 files changed, 121 insertions(+), 16 deletions(-) create mode 100644 knowledge base/uv.md diff --git a/knowledge base/ai/agent.md b/knowledge base/ai/agent.md index f8df09d..0851afd 100644 --- a/knowledge base/ai/agent.md +++ b/knowledge base/ai/agent.md @@ -1,12 +1,10 @@ # AI agent -> [!caution] -> TODO - -AI-enabled system or application capable of autonomously performing tasks of various complexity levels on their own, -possibly **without** the need to stop to ask permission or consent to the user. +AI-enabled system or application capable of _autonomously_ performing tasks of various complexity levels by designing +workflows and using the tools made available to them. 1. [TL;DR](#tldr) +1. [Skills](#skills) 1. [Concerns](#concerns) 1. [How much context is too much?](#how-much-context-is-too-much) 1. [Security](#security) @@ -17,10 +15,14 @@ possibly **without** the need to stop to ask permission or consent to the user. ## TL;DR -Agents design their workflow and utilize the tools that are made available to them.
-They use natural language processing techniques of [LLMs][large language model] to comprehend user inputs, respond to -them step-by-step, and determine when to call on external tools to obtain up-to-date information, optimize workflows -and create subtasks autonomously to achieve complex goals. +AI agents can encompass a wide range of functions beyond natural language processing.
+These functions include making decision, problem-solving, interacting with external environments, and performing +actions. + +Agents design their own workflow and utilize the tools that are made available to them.
+They use [LLMs][large language model] to comprehend user inputs, deconstruct and respond to requests step-by-step, +determine when to call on external tools to obtain up-to-date information, optimize workflows, and autonomously create +subtasks to achieve complex goals. Traditional software is _deterministic_, AI is _probabilistic_. @@ -44,6 +46,13 @@ them during the run. Prefer **requiring** consent by agents when running them. +## Skills + +Skills extend AI agent capabilities with specialized knowledge and workflow definitions. + +[Agent Skills] is an open standard for skills. It defines them as folders of instructions, scripts, and resources that +agents can discover and use to do things more accurately and efficiently. + ## Concerns Agents created by Anthropic and other companies have a history of not caring about agent abuse, and leave users on @@ -143,6 +152,7 @@ See [An AI Agent Published a Hit Piece on Me] by Scott Shambaugh. [39C3 - Agentic ProbLLMs: Exploiting AI Computer-Use and Coding Agents]: https://www.youtube.com/watch?v=8pbz5y7_WkM [39C3 - AI Agent, AI Spy]: https://www.youtube.com/watch?v=0ANECpNdt-4 +[Agent Skills]: https://agentskills.io/ [Agentic ProbLLMs - The Month of AI Bugs]: https://monthofaibugs.com/ [AI Doesn't Reduce Work — It Intensifies It]: https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it [An AI Agent Published a Hit Piece on Me]: https://theshamblog.com/an-ai-agent-published-a-hit-piece-on-me/ diff --git a/knowledge base/ai/claude/claude code.md b/knowledge base/ai/claude/claude code.md index 6c976db..7c17c47 100644 --- a/knowledge base/ai/claude/claude code.md +++ b/knowledge base/ai/claude/claude code.md @@ -203,9 +203,13 @@ Manually add the MCP server definition to `$HOME/.claude.json`: ## Using skills -Refer [Skills][documentation/skills]. +Refer [Skills][documentation/skills].
+See also: -See also [create custom skills] and [Prat011/awesome-llm-skills]. +- [Create custom skills]. +- [Prat011/awesome-llm-skills]. + +Claude Skills follow and extend the [Agent Skills] standard format. Skills superseded commands.
Existing `.claude/commands/` files will currently still work, but skills with the same name will take precedence. @@ -308,6 +312,7 @@ Claude Code version: `v2.1.41`.
- [Gemini CLI] - [OpenCode] - [Prat011/awesome-llm-skills] +- [Claude Skills vs. MCP: A Technical Comparison for AI Workflows] ### Sources @@ -337,7 +342,9 @@ Claude Code version: `v2.1.41`.
[Website]: https://claude.com/product/overview +[Agent Skills]: https://agentskills.io/ [AWS API MCP Server]: https://github.com/awslabs/mcp/tree/main/src/aws-api-mcp-server +[Claude Skills vs. MCP: A Technical Comparison for AI Workflows]: https://intuitionlabs.ai/articles/claude-skills-vs-mcp [Cost Explorer MCP Server]: https://github.com/awslabs/mcp/tree/main/src/cost-explorer-mcp-server [pffigueiredo/claude-code-sheet.md]: https://gist.github.com/pffigueiredo/252bac8c731f7e8a2fc268c8a965a963 [Prat011/awesome-llm-skills]: https://github.com/Prat011/awesome-llm-skills diff --git a/knowledge base/ai/llama.cpp.md b/knowledge base/ai/llama.cpp.md index f5f3038..de028e5 100644 --- a/knowledge base/ai/llama.cpp.md +++ b/knowledge base/ai/llama.cpp.md @@ -41,7 +41,7 @@ llama-cli -m 'path/to/target/model.gguf' -md 'path/to/draft/model.gguf' # Download and run models. llama-cli -mu 'https://example.org/some/model' # URL llama-cli -hf 'ggml-org/gemma-3-1b-it-GGUF' -c '32.768' # Hugging Face -llama-cli -dr 'ai/qwen2.5-coder' --offline # Docker Hub +llama-cli -dr 'ai/qwen2.5' --offline # Docker Hub # Launch the OpenAI-compatible API server. llama-server -m 'path/to/model.gguf' diff --git a/knowledge base/ai/llm.md b/knowledge base/ai/llm.md index 3de400f..a6fb8c0 100644 --- a/knowledge base/ai/llm.md +++ b/knowledge base/ai/llm.md @@ -133,11 +133,17 @@ is correct by breaking questions in smaller, more manageable steps, and solving final answer.
The result is more accurate, but it costs more tokens and requires a bigger context window. -The _ReAct loop_ (reason+act) forces models to loop over chain of thoughts.
-A model breaks the request in smaller steps, acts on those using [functions][function calling] if they deem it useful, -checks the results, updates the chain of thoughts, and repeat until the request is satisfied. +The _ReAct loop_ (Reason + Act) paradigm forces models to loop over chain-of-thoughts.
+A model breaks the request in smaller steps, plans the next action, acts on it using [functions][function calling] +should it decide it needs to, checks the results, updates the chain of thoughts, and repeats this Think-Act-Observe loop +to iteratively improve upon responses. -Next step is [agentic AI][agent]. +The _ReWOO_ (Reasoning WithOut Observation) method eliminates the dependence on tool outputs for action planning.
+Models plan upfront, and avoid redundant usage of tools by anticipating which tools to use upon receiving the initial +prompt from the user.
+Users can confirm the plan **before** the model executes it. + +[AI agents][agent] use these methods to act autonomously. ## Prompting @@ -195,6 +201,7 @@ Refer: ## Further readings - [SEQUOIA: Serving exact Llama2-70B on an RTX4090 with half-second per token latency] +- [Optimizing LLMs for Performance and Accuracy with Post-Training Quantization] ### Sources @@ -203,6 +210,7 @@ Refer: - [Local LLM Hosting: Complete 2026 Guide - Ollama, vLLM, LocalAI, Jan, LM Studio & More] - [LLM skills every AI engineer must know] - [Function calling in LLMs] +- [What is chain of thought (CoT) prompting?] + +## Further readings + +- [Website] +- [Codebase] + +### Sources + +- [Documentation] + + + + + + + +[Codebase]: https://github.com/astral-sh/uv +[Documentation]: https://docs.astral.sh/uv/ +[Website]: https://docs.astral.sh/uv/ + +