agents-workspace — a workflow layer that orchestrates The Agency agents intelligently.

[wcgomes]

Thank you, agency-agents community! 🙌

I created agents-workspace — a specialist-first, subagent-driven workflow layer that enforces intelligent task delegation across your 144+ agents.

The workflow difference

Instead of hoping the right agent picks up the right context (or specifying it yourself):

Before	After
Raw prompt → single agent	Task analyzed → specialist selected via delegate
Context window bloated after 3 turns	Each subagent gets focused, isolated context
All tasks sequential	Independent tasks run in parallel
Repeat the same mistakes	Memory persists via wiki
No validation before coding	implement skill validates understanding + enforces minimal changes
Changes spread across files	Minimal, targeted diffs per task
Same pattern, same fix every time	agents-skills evolves local skills from 3x repeated patterns

How it works

1. Task arrives → Load delegate to assess: define done criteria, consult wiki for existing workspace context
2. Specialist discovery → delegate scans available agents, selects the best semantic match per task scope/constraints
3. Execution → Delegate to specialists, parallelize non-overlapping tasks, enforce implement skill for all code changes
4. Error handling → debug skill automatically triggers for unexpected behavior/errors
5. After completion → wiki ingests learnings, lints for stale references or contradictions
6. Pattern detected 3x? → agents-skills proposes or creates a local skill to automate the workflow
7. Closure → Review all delegated output against AGENTS.md rules before marking complete

What you get

5 core orchestration skills plus your full agent pool, enforced by mandatory AGENTS.md rules:

• delegate — Mandatory specialist discovery, selection, and delegation workflow
• wiki — Persistent workspace memory for pre/post task context
• implement — Pre-change validation, minimal diffs, and code correctness enforcement
• debug — Error investigation and unexpected behavior resolution
• agents-skills — Local skill creation, refinement, and validation from repeated patterns
• Core rules: brevity, specialist-first fallback, subdelegation governance, invocation contracts

Quick start

curl -sL https://raw.githubusercontent.com/wcgomes/agents-workspace/main/tools/install.sh | bash
cp AGENTS.md /your-project/

Tell your agent: "Set up the wiki and start working"

Devcontainer fans 🎉

For those who prefer devcontainers, I also created dedicated features:

• agents-workspace
• agency-agents

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!

GitHub

[Wuzoobia]

Excellent!
Thank you!