Agentic

Agentic is a minimal queue-first project skeleton for experimenting with isolated AI development agents.

The core idea is simple:

• agents do not share conversation context
• tasks move through .queue/pending, .queue/in-progress, .queue/done, and .queue/failed
• role files under agents/ define responsibilities
• helper scripts under scripts/ validate and move tasks
• docs record the operating model, hardening rules, and review/test/ops workflows

This repository is the public minimal core. Product-specific automation, private dashboards, long-running agents, and organization-specific memory should live in a private derived repository.

Quick Start

python3 scripts/control_check.py
python3 scripts/enqueue.py docs/tasks/example.architect-task.json
python3 scripts/claim_task.py task-0001 --agent local-architect --role architect
python3 scripts/complete_task.py task-0001 --summary "Reviewed example task" --validation "manual smoke check"

CLI Smoke Test

python3 -m agentic.cli

Tests

pytest
bash tests/test_queue_status.sh
python3 scripts/control_check.py
python3 scripts/public_safety_check.py

Project Layout

• agents/: role instructions for Architect, Coder, Reviewer, Tester, and Ops
• .queue/: file-based task lifecycle
• scripts/: queue and control helpers
• docs/: public-safe architecture, hardening, workflow, and checklist docs
• tests/: focused smoke and queue helper tests

Public Safety

Before pushing changes to public core, run the public release checklist in docs/checklists/public-release-checklist.md.

Status

This is an intentionally small core. Keep it generic, dependency-light, and safe to fork.