Large-Language-Model on Janik von Rotz

Working with LLM agents

Tue, 17 Feb 2026 16:17:07 +0100

In the last few days I tinkered with Cortecs and OpenCode. It seems LLM agents are here to stay and I need to deal with them. I admit that I am very skeptic of the AI hype. However, I see a lot of benefits when working with LLM agents in a controlled environment.

People have built great tools and practices. There are many providers that host LLMs or route between multiple providers. The ecosystem as matured and this led me to explore more and put in more effort to understand what is going on.

The goal of this exploration was to find a new coding workflow that includes an LLM agent and fits well into my existing workflow.

As the good engineer that I am, I first wrote down some requirements for the workflow.

Requirements

No unnecessary features

The workflow should not encourage me to build more stuff, but to fix issues and improve features.

Code is technical debt and the theory of code is cognitive debt. Building a lot of features and not using them is producing waste and noise.

No digressions

In my experience, if the agent does not satisfy the requirements, you start explaining the issue over and over again. You often end up in argument and LLM always promises to do better.

This kind of arguing and guessing can be a huge waste of time and defeats the very reason for using an agent. Instead of having an ongoing conversation and a bloated context, it is often easier to start over.

And thus the workflow should encourage me to abort and refine my initial prompt.

Spec-driven prompts

Form experience it is worth to write an elaborate prompt. The initial prompt is a specification of of the new feature or bug to resolve.

Before starting an agent session, a well defined prompt must already exist.

Quick boostrap

I have many software projects with a different layout and framework. Before letting an agent go rough in the repo, I want to provide clear guideliens.

The workflow should make it easy to provide the necessary context for very different projects.

Task file integration

This is the most important requirement. I manage all my software projects with taskfile.build. It is a standard for a bash script and also a library for commands. It allows me to manage and bootstrap very different software projects.

Whatever workflow I am going to build, it can be called from a task file.

Workflow

Now let me show you how the coding workflow looks like.

After iterating on many ideas, practices, scripts and commands, I ended up with three step workflow:

Setup: Prepare the project to run an LLM agent
Run: Create and run a prompt
Finish: Summarize and commit the changes

The commands I use in my workflow are imported from https://taskfile.build/library/#llm.

Setup

First, I enter an new or existing project and run the init project command:

task init-project

This command creates several files and folders. The relevant files are:

AGENTS.md: Tempalted guideline for the LLM agent
README.md: Guide for the user and the LLM agent
prompts: Folder that contains spec-driven prompts
task: Templated project-level task file

I update these files according to the context of the project.

Run

I create new prompt file with this command:

task create-prompt "Add prompt commands"

This creates and opens the file prompts/08_add-prompt-commands.md in the default editor. Then I update the task section with the specification.

Here is an example of what the prompt file might look like before being executed:

---
title: Add prompt commands
---

# Run 08

Replace the `==` marked instructions in this file while you work on the task.

## Context

Read the `AGENTS.md` and `README.md` to get understanding of the project.

## Task

Add two new commands `create-prompt` and `list-prompt` to the `bin` folder.

The first command asks for a title. The user is expected to enter something like `Title of the prompt`.
Then the command creates a file `prompts/NN_title-of-the-prompt.md` from a template `prompt.md.template`. It writes the title of the prompt into the frontmatter.
Once the file is created it uses the `$EDITOR` to open the file.
The number sequence starts at 01 and continues upwards.

Similar to `list-dotenv` the `list-prompt` commands lists all prompts. It shows a simple table:

```markdown
| ID  | title               |
| --- | ------------------- |
| 08  | Add prompt commands |
```

The title is retrieved from the frontmatter. The table width should be dynamic.

Add the new commands to the `library.md` in the LLM section. Update the `task.template` with `task update-template`.

## Worklog

==Fill this in as you work on the task==

## Summary

==Fill this once you completed the task==

In order to execute the prompt I launch opencode in the root of the project and simply copy-paste the path to the prompt file. Open Code then starts working on the task.

If it does not get the instructions, I either create a new session or give some manual inputs.

Finish

After a few seconds the task has been completed. The worklog and summary section should completed by the agent.

It is crucial to not blindly trust the completeness of your spec or the output of the LLM. So I check the git diff. If somethings seems off, I either update the prompt or make manual inputs.

If I am happy with the result, the changes are staged gaa and I run task update-changelog-with-llm. This will update the CHANGELOG.md according to the keep a changelog specification.

To commit the changes I use task commit-with-llm. This command generates the commit message from the git diff according to the Conventional Commits specification.

And that’s it!

Afterthoughts

This workflow works well for trivial and well defined tasks. Once you try more complex task and multiple agents, the quality of the project erodes pretty fast.

Some researchers have demonstrated that LLMs can create entire browsers within a week. Browser engines are well specified, but their creation takes a very long time. The result met all requirements, but was very slow and inefficient.

I believe that the AI hype is a challenge for the entire software industry. It’s not just salespeople who are hyping AI, but also engineers who should know better.

Until the bubble pops, I’ll try my best.

Work with LLMs on the command line

Mon, 27 Jan 2025 16:28:35 +0100

I used VSCodium and Codeium to develop Python code. VSCode is my editor of choice and Codeium is a well integrated AI-tool that helps writing code. While it solved a lot of problems for me, especially writing boilerplate code, I became more and more frustrated using this setup.

No longer I wanted to understand the actual problem or piece of code, but just to prompt out a solution. Often I was eagerly waiting for the auto-complete feature to fix my code. I copied pieces of code to LLM chats in the browser and then updated the code in the editor. This workflow didn’t feel right. This isn’t coding.

The obvious solution would be to to stop using these AI-tools and write code as I used to. On one side I really would like to do that, but on the other side, I feel pressured to use this technology. I believe in the LLM technology and the promise of more efficient coding. However, I do not like the products and services built around LLMs.

This is also what annoyed me about Codeium. Every prompt is funneled through their service. If they change the terms of agreement the only option is to agree. Coding should not be on a providers agreement. Especially if the terms are to their favor.

Inspiration

So I was looking for different angles on this topic. I was not looking for alternatives. Alternatives often don’t solve the actual problem. While reading Hacker News (HN) I got an inspiration. In comment section of Yek: Serialize your code repo (or part of it) to feed into any LLM I stumbled this:

https://github.com/simonw/files-to-prompt/ seems to work fine especially for Claude as it follows the recommendations and format.

In the HN sphere Simon Willison (guy in the picture) is well-known for his blog and insights. He writes a column on how he uses LLMs to code and most importantly is he the author of the llm command line interface (cli).

The link in the command shows an example on how Simon feeds multiple files into an LLM using his cli. The llm cli allows he to switch between remote models and even use local LLMs! There is a well maintained plugin directory to attach any kind of models.

Files to prompt

So I started to introduce the files-to-prompt solution into my development workflow. My worklow is heavily based on taskfile.build. For every project there is a task script file to run selected commands.

For the Odoo.Build project I added a new command:

printf "| %-${cmd_width}s | %-${opt_width}s | %-${desc_width}s |\n" "update-with-llm" "[path][prompt]" "Update files with LLM prompt."

And here is the function that does the magic:

function update-with-llm() {
    if test -z "$1"; then 
        echo "\$1 is empty."; 
        exit 1; 
    fi

    # Get files from path
    if [ -f "$1" ]; then
        FILES="$1"
    elif [ -d "$1" ]; then
        FILES=$(find "$1" -type f \( -name "*.md" -o -name "*.yml" -o -name "*.yaml" -o -name "*.conf" -o -name "*.nginx" \) \
            | grep -vFf <(git -C "$1" ls-files --ignored --exclude-standard --others))
    else
        echo "$1 is neither a file nor a directory."
        exit 1
    fi
    echo -e "Loaded these files into prompt:\n\n$FILES\n"

    # Prompt task description
    if [ -z "$2" ]; then
        read -p "Enter the task description: " TASK_DESCRIPTION
    else
        TASK_DESCRIPTION="$2"
    fi

    # Prepare the files content for prompt
    FILE_CONTENTS=""
    for FILE in $FILES; do
        FILE_CONTENTS+="<<<$FILE>>>
$(cat "$FILE")

"
    done

    # Define prompt content
    mkdir -p tmp
    PROMPT_FILE="tmp/llm_update"
    echo -e "\nWrite prompt to $PROMPT_FILE."
    cat << EOF > "$PROMPT_FILE"
Look at the code files below and do the following:

$TASK_DESCRIPTION

Output all files that you need to change in full again, including your changes. 
In the same format as I provide the files below. Under no circumstances output any other text, 
no additional infos, no code formatting chars. Only the code in the given format.

Here are the files:

$FILE_CONTENTS
EOF

    # Run the llm command
    echo -e "Send prompt and wait for the response of the $LLM_MODEL LLM."
    RESULT=$(cat "$PROMPT_FILE" | llm -m "$LLM_MODEL")

    # Check if result is empty
    if [ -z "$RESULT" ]; then
        echo "No response from the model. Exiting."
        exit 1
    fi

    # Save the result to a file
    RESULT_FILE="tmp/llm_update_result"
    echo "$RESULT" > "$RESULT_FILE"
    echo -e "Saved response to $RESULT_FILE.\n"

    # Show a preview of result file
    less "$RESULT_FILE"

    # Ask for confirmation before updating files
    read -p "Do you want to apply these updates to the files? (y/n): " CONFIRM
    if [[ ! "$CONFIRM" =~ ^[Yy]$ ]]; then
        exit 0
    fi

    # Parse the response from the file and update the files directly
    echo -e "Parsing the response and updating files...\n"
    CURRENT_FILE=""
    while IFS= read -r LINE; do
        if [[ $LINE =~ ^"<<<"(.*)">>>" ]]; then
            CURRENT_FILE="${BASH_REMATCH[1]}"
            echo "Update file $CURRENT_FILE."
            > "$CURRENT_FILE"
        elif [[ -n $CURRENT_FILE ]]; then
            echo "$LINE" >> "$CURRENT_FILE"
        fi
    done < "$RESULT_FILE"
}

Executing the function does:

Collect all Python and XML files from the provided path.
Ask for task instructions.
Generate a prompt with the instruction and files that asks for a response with file content only.
Pass the prompt to the llm command.
Store the response into a file.
Show a preview of the file with less.
Ask if the changes should be applied.
Apply the changes on confirmation.

In Action

Let me show the function in action. In this scenario I have an Odoo module board_user_acl and would like to rename a definition.

First I run the task file with the update-with-llm command and give the path to the module as parameter:

task update-with-llm addons/server_tools/board_user_acl "Rename restricted to restrict"
Loaded these files into prompt:

addons/server_tools/board_user_acl/__manifest__.py
addons/server_tools/board_user_acl/security/security.xml
addons/server_tools/board_user_acl/views/menu.xml

The function asks for a task description:

Enter the task description: Rename restricted to restrict

And then it waits for the response.

Write prompt to tmp/llm_update.
Send prompt and wait for the response of the gpt-4o LLM.
Saved response to tmp/llm_update_result.

Once its done, a preview of the updated files is shown:

<<<addons/server_tools/board_user_acl/__manifest__.py>>>
{
    "name": "Board User ACL",
    "summary": """
        Restrict access to dashboards app.
    """,
    "author": "Mint System GmbH",
    "website": "https://www.mint-system.ch",
    "category": "Technical",
    "version": "16.0.1.0.0",
    "license": "AGPL-3",
    "depends": ["base_user_acl"],
    "data": ["security/security.xml", "views/menu.xml"],
    "installable": True,
    "application": False,
    "auto_install": False,
    "images": ["images/screen.png"],
}
tmp/llm_update_result (END)

Quit the less preview with q and confirm the change.

Parsing the response and updating files...

Update file addons/server_tools/board_user_acl/__manifest__.py.

That’s it.

I will most likely add new commands and other prompts.

Edits:

Removed prompt response with git patch file as it was often not working correctly.
Renamed llm-update to update-with-llm
Updated how files are searched