agents-workspace — A lightweight agent workflow with SDD and OKF concepts powered by The Agency.
Hello agency-agents community!
I created agents-workspace with the ideia to create a professional workflow with dynamic team assembly of specialized agents tailored to each task, parallel execution, optimized context windows, and a self-learning cycle.
Under the hood, its operating model brings together memory and OKF concepts, subagent-driven development, spec-driven workflows, and agent loops grounded in empirical verification.
I added the agents from this repository as the default agents catalog on the installer. The user can install all agents or select the desired division.
It's the workflow we use everyday at work from software development to research, marketing etc.
User request
|
v
+--------------------------------------+
| Main agent (coordinator) |
| Read wiki/index.md for context |
| Optional: use available knowledge |
| tools (e.g. code/symbol/dependency |
| graphs). |
+--------------------------------------+
|
+--> [when a durable behavior contract is needed]
| Load spec-builder; confirm spec workflow
|
v
+--------------------------------------+
| Load orchestrate; assemble and plan |
| Optional: delegate spec drafting and |
| planning -> specs/... -> Spec ref |
| Delegate execution using Spec ref |
| when present |
+--------------------------------------+
|
| Illustrative/default pattern for independent scopes:
| orchestration may use one agent or sequential handoffs instead.
|
+----------------+----------------+----------------+
v v v
+------------+ +------------+ +------------+
| Specialist | | Specialist | | Specialist |
| scoped work| | scoped work| | scoped work|
+------------+ +------------+ +------------+
\ | /
\ | /
+--------------+--------------+
v
+----------------------+
| Merge deliverables |
+----------------------+
|
v
+----------------------+
| Review / verification|
+----------------------+
|
+--------+--------+
| |
| done criteria | changes needed?
| met |
v v
| +----------------------+
| | Coordinator sends |
| | findings to affected |
| | specialist(s) |
| +----------------------+
| |
| v
| +----------------------+
| | Re-dispatch / revise |
| +----------------------+
| |
| v
| +----------------------+
| | Integrate adjustments|
| +----------------------+
| |
| +--> Review / verification again
v
+----------------------+
| wiki: post-review |
| learning evaluation |
+----------------------+
|
v
+----------------------+
| Uncaptured durable |
| workspace knowledge |
| remains? |
+----------------------+
|
+--------+------------+
| |
| no | yes
v v
+----------------+ +----------------------+
| No ingestion | | One consolidated |
| dispatch | | serialized ingestion |
| | | stream |
+----------------+ +----------------------+
| |
| +----------------------+
| | Coordinator reviews |
| | ingestion result |
| +----------------------+
| |
+----------+----------+
v
+----------------------+
| Synthesis |
| -> final response |
+----------------------+
Devcontainer fans
For those who prefer devcontainers, I also created dedicated features:
Add them to your devcontainer.json and get everything pre-configured in your container.
Install once, use it across every project.
Feedback and PRs welcome!
Source: msitarzewski/agency-agents