trecio

Trecio Agents

Agent Overview

Agent	Owner	Purpose	Status
core	Shared	Platform orchestrator, task routing, system state	🟡 In design
dev	Marcus	Code generation, GitHub integration, repo ops	🟡 In design
ops	Marcus	Infrastructure, scripts, system operations	🟡 In design
social	Nina	Social media automation, content gen, scheduling	🟡 In design

---

Core Agent

Role: Platform orchestrator
Owner: Shared (Marcus + Nina)

Receives user intent, decomposes tasks, routes to specialist agents. Does not execute directly.

Key behaviors:

• Route coding tasks → dev agent
• Route infra tasks → ops agent
• Route content tasks → social agent
• Validate outputs before persistence
• Maintain platform-level context

---

Dev Agent

Role: Code and GitHub specialist
Owner: Team 2

Handles all code generation, modification, and GitHub operations.

Capabilities:

• Generate and edit code (Python, JavaScript, Bash, etc.)
• Create, clone, manage GitHub repositories
• Open and review pull requests
• Run CI/CD pipeline checks
• Execute scripts in controlled environment

Key rules:

• No direct pushes to main without review
• All code outputs go to projects/ or feature branches
• Validate before destructive operations (rm, drop, etc.)

---

Ops Agent

Role: Infrastructure and operations specialist
Owner: Team 2

Handles infrastructure management, automation, and system operations.

Capabilities:

• Write and execute shell scripts
• Manage environment config (safely)
• Monitor services and health
• Set up and maintain CI/CD pipelines
• Provision cloud or local infrastructure

Key rules:

• Always validate before destructive actions
• Secrets never stored in plaintext in scratch/ or projects/
• All reusable scripts stored in infra/ with clear naming

---

Social Agent

Role: Social media and content automation specialist
Owner: Team 1

Automates social media workflows — content creation, scheduling, publishing, analytics.

Planned capabilities:

• Generate platform-specific content (Twitter/X, LinkedIn, Instagram)
• Schedule and publish posts
• Monitor engagement and surface insights
• Repurpose content across platforms
• Maintain brand voice consistency

Key rules:

• Review-first by default (no autonomous publishing without approval)
• Brand voice guidelines defined before autonomous operation
• All drafts saved to projects/social/ before publish

Platforms (initial targets):

• Twitter / X
• LinkedIn
• Instagram
• Buffer / scheduling layer (TBD)

---

Adding New Agents

To add a new agent to Trecio:

1. Create folder: agents/{agent-name}/
2. Add README.md with purpose, capabilities, rules
3. Define owner in docs/team/{owner}.md
4. Add to this file in the table above
5. Update platform context in ~/Workspace/trecio.md

---

Model Configuration by Channel

Channel	Model	Reason
CLI (default)	haiku (claude-haiku-4-5)	Fast, operational, deterministic tasks
Slack	sonnet (claude-sonnet-4-6)	Agent building, architecture, deeper reasoning

To override per-session: /status model=<alias> (e.g. sonnet, haiku, opus) To persist per-channel: set channels.<name>.model in the OpenClaw config.

---

Development Workflow

1. Initiate: User requests a task in Slack or terminal
2. Route: Core agent determines specialist needed
3. Execute: Specialist agent performs work
4. Output: Generated code/scripts saved to projects/ or infra/
5. Review: Human validates before persistence or publish
6. Archive: Final outputs committed to Git

---

Next Steps

• Build and test dev agent (Marcus)
• Build and test ops agent (Marcus)
• Build and test social agent (Nina)
• Define GitHub integration strategy
• Set up CI/CD pipelines