Code Lab
Agency Agents Explained: Role-Based Instructions for AI Workflows
CODE LAB · PROJECT SPOTLIGHT
Agency Agents Markdown · MIT
Explore Agency Agents’ specialist profiles, installation choices and actual Markdown source, including what role prompts can and cannot guarantee.
Preview: msitarzewski/agency-agents on GitHub
An AI assistant often needs more than a job title. It needs a concrete brief: which problems to focus on, what to deliver and how the result will be reviewed. Agency Agents collects specialist profiles that make those expectations explicit.
The repository groups agent definitions across engineering, design, marketing and other disciplines. These definitions describe a role, working style, responsibilities and outputs. The README also points to a separate desktop application for browsing and installing profiles. The instruction files and that application are distinct pieces of the ecosystem.
This source-and-documentation review was checked November 13, 2026. Project-supplied images and reported results are attributed below; no new hands-on benchmark is claimed.
What does Agency Agents do?
For a frontend task, choose a relevant profile and provide the actual product requirements, repository conventions and acceptance criteria. Ask for a bounded result, such as improving keyboard navigation in one component. A specialist profile can focus attention, but it still needs access to the right files and a clear definition of completion.
Inside the repository: a frontend profile
This excerpt comes from the frontend developer profile. It records the intended role and personality in Markdown. It is source material for an agent’s instructions, rather than an executable implementation or proof of the model’s expertise. The full file expands these ideas into responsibilities such as accessibility and performance review.
Source: msitarzewski/agency-agents at 8ae6164. License: MIT. Original lines retained; syntax colors and line numbers added for this article.
Source credit and MIT license notice
MIT License Copyright (c) 2025 AgentLand Contributors Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Getting started with a bounded task
Read the selected profile before installing it. The project documents script-based installation and a separate application; choose the path supported by your coding client. Install a small set first so overlapping roles do not create confusing instructions. Keep a record of the profile revision when comparing results across sessions.
Limits and what to verify
The catalog’s descriptions are project claims about intended behavior. A persona does not add knowledge, permissions or reliable memory by itself. Those capabilities depend on the host agent and its tools. Review generated code and factual claims using the same standards you would apply without the profile, and resolve conflicts with your project’s own instructions.
Where it fits in your workflow
The useful experiment is to compare a normal task brief with a carefully chosen profile plus the same brief. Look for clearer outputs and fewer missed requirements. For a related interface that runs coding agents, see T3 Code; Agency Agents contributes role instructions that a compatible host can consume.
Related Code Lab: T3 Code and AI coding-agent interfaces. For the source-review format used here, see our Lichii repository walkthrough.
Official references: Repository · README at reviewed revision · Release history · Source excerpt.