AI Agents System
The project utilizes a multi-agent system to automate development, documentation, and content workflows.
Agents Overview
| Agent | Role | Definition Source |
|---|---|---|
| Docser | Unified Documentation, History, and Sync Agent. (Replaces Secretary & Saver). | .github/agents/Docser.agent.md |
| Kant | Research assistant & Content generator (Immanuel Kant persona). | .github/agents/kant.agent.md |
| Writer | WordPress content manager via PHP CLI scripts. | .github/agents/Writer.agent.md |
| Tester | QA agent for robust local and CI Playwright UI verification. | .github/agents/Tester.agent.md |
| Hubber | GitHub Project manager tracking tasks and backlogs. | .github/agents/Hubber.agent.md |
| RAG Service | Provides semantic search and documentation management via MCP. | Docker Container |
Infrastructure
Synchronization
Agents are defined in the repository (specifically the ai-tabs branch) and synchronized to the server automatically.
- Script:
scripts/pull_agents_cron.sh - Source:
origin/ai-tabs - Target:
.github/agents/ - Schedule: Runs every 5 minutes via cron.
# Manual Sync Command
/root/sites/vania-novikau.me/scripts/pull_agents_cron.sh
General Scripting Protocol
In order to prevent root folder clutter, all AI Agents across the project have been hard-coded with strict instructions instructing them on how to manage any diagnostic, deployment, analysis, or operational scripts.
- Workspace Path: Agents must NEVER save temporary files, code updates, or dumps inside the root repository
vania-novikau.me/. - Target Location: Scripts must be placed inside the
<local_root>/ai-scripts/directory.
Agent Skills System
We have modernized the scripting protocol to transition towards Agent Skills (following the open standard by AgentSkills.io). Instead of flat scripts, AI capabilities are bundled into standalone folders inside agent-skills/ with accompanying SKILL.md metadata, allowing for Progressive Disclosure and reusable multi-step workflows.