Browse / Second brain / agent-project-template
agent-project-template
Yokhan
Self-deploying AI agent workspace template: 8 hooks, 7 agents, 11 skills, 8 commands, Obsidian brain, self-improvement loop.
Open Yokhan/agent-project-template → Checked against the four rules.
Works with any agent
ships AGENTS.md, the cross-tool instruction file.
In its own words
Based on 20+ production repos.
Self-deploying AI-agent optimized project template with MCP-based dynamic rule routing, a pinned stack-aware code-intelligence toolchain, Codex repo-scoped skills and subagents, persistent memory, autonomous work loops, self-improvement, and merge-safe sync. Token savings are benchmarked per project instead of assumed globally.
Подробная инструкция на русском: SETUP_GUIDE.md — пошаговая настройка, MCP-серверы, Zed, troubleshooting. Product boundary: docs/PRODUCT_BOUNDARY.md • Safe defaults: docs/SAFE_DEFAULTS.md • Supported environments: docs/SUPPORTED_ENVIRONMENTS.md • Codex fan-out: docs/CODEX_FANOUT_PATTERNS.md • Template releases: docs/TEMPLATE_RELEASES.md
If An Agent Only Has This GitHub Link
Follow the canonical update protocol:
- Classify source, generated downstream, or legacy downstream; never sync the source into itself.
- Read installed version from
.template-manifest.json. - Explicit user/AgentOS tag wins. Only when no tag was selected, resolve the current stable release at .
- Verify
git remote get-url template; never silently replace a conflict. - Run pinned dry-run, then apply the same tag. Bare
--from-gitis canary-only. - Use the target release checkout's script with
--project-dirwhen local sync is stale. - Verify manifest version, diff, overlays, conflicts, and checks before success.
The repository's own description, then 13% of its README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.