Notable verbatim quotes

⚠️ Speaker labels are absent in the source. Attributions below are inferred from context — treat with the hedging rules in SKILL.md.

On the "dark factory" / orchestrator framing

• [dark factory framing, §1] "At Tessl, we've been trying to turn our own software development lifecycle into full dark factory ... We've been focusing on the part of the process that is linear issue to merge PR. And you might call this like an orchestrator."

On mega-prompts and their smells

• [prompt theater, §4] "You were a seasoned world-class senior quality engineer who's 20 plus years hands on ... These don't really work. These are a little bit of theater."
• [begging, §4] "Usually like all caps is kind of a smell for begging. It's like please do. I capitalize it so I'm serious."
• [no branching, §4] "If you know about how LLMs process input. They don't do branching. When it comes to like imbibing a prompt, it's all or nothing. It's going to read all these tokens or none of them."

On skills

• [skills frontmatter, §8] "What is so special is a tiny bit on top of this prompt which is called the front mechanism from leather it's a yaml piece inside markdown file ... this YAML part has two items. The first is name and the second is description."
• [description is crucial, §8] "The name is the name, but description is description is where the magic happens. Our agents are going to preload all the available skills for them, but only the discriminators, only the inscriptions. And then we'll apply logic or reasoning ... the description is absolutely crucial."
• [conditional activation, §8] "This conditional activation for effectively like little reusable atomic prompts was kind of a game changer."
• [Claude Code activation, §9] "Claude Code was actually one of the worst activators of skills. So no matter what you put in the prescription. No matter what."

On the deterministic / non-deterministic split

• [the heuristic, §11] "Every time we want something predictable we walk away from an llm and convert it to script."
• [scripts are free, §11] "Scripts are direct invocations zero tokens zero money scripts are awesome."
• [agents over-scripting, §11] "Agents love regex it's like it's their go-to tool for everything ... it will create stupid lists of allowed words and blocked words and then write absolutely crazy regexes ... this is obviously wrong thing to do because we want the reasoning to be applying to reading text."
• [Symphonia anecdote, §11] "This is the story of how I spent 60 million tokens trying to check if the PR was open or closed."

On rules

• [what rules are, §13] "Those things that should not be missed can be put in another primitive that it's called rules. Are tiny prompts. That are going to be there at every time a conversation is sent to the model. No matter what's going on."
• [skills vs rules cost, §13] "The only part [of a skill] that doesn't get loaded every time is the actual body of the skill the description will be loaded every time so that's also mandatory tokens."
• [Claude doesn't natively support rules, §13] "Claude doesn't have native sports rules that's the hack that we do for Claude 24 souls we put it in .claude and point rules."

On dumb models

• [Haiku is enough, §12] "We don't have like I said we don't have genius level problems we just don't we like to think we do ask your therapist about it really don't ... Haiku can probably do most of what you're doing."
• [OpenRouter / LiteLLM, §12] "There are at least two great projects out there. The one is called open router ... the other is called liteLLM that allow you to run the free models for almost free."

On context as artifact

• [sources vs artifacts, §14] "Please keep your skill sources, script sources and rule sources in github where you tag them when you package them use registry and then you will have a distributable piece of software that works just fine."
• [the JAR analogy, §14] "Who commits jar [files] to github? Thank you whoever don't do that won't do it in this context artifact as well."

On evals

• [evals not tests, §15] "We run the same question without any additional context. And we run the same question then prompt basically with our [plugin] example. And we compare the results if the results are better with context that means that yay we did a good context."
• [LLM-as-judge, §15] "Is it generally better. Most of the time? And that's all we can hope for for non-deterministic miss code."

Closing meta-point

• [policy as plugin, §16] "We can write skill that will tell us. Analyze our prompts. And if you see something that should be rule. Make it rule. And something that should be scriptable [make it a] script ... install it in your CI. And the knowledge of this workshop will become policy that will be enforced on your code."