Agentic Engineering Workflow for Principal Engineers
source post: L8 Principal's Agentic Engineering Workflow
L8 Principal's Agentic Engineering Workflow
Source: youtube · Kun Chen Saved: 20260620 Tags: youtube Display: Agentic Engineering Workflow for Principal Engineers — A terminal-centered AI dev harness combining memory files, reusable skills, planning artifacts, and validation gates for agent output.
TL;DR
Kun Chen's L8 Principal agentic engineering workflow is a terminal-centered AI development harness that combines memory files, reusable skills, interactive planning artifacts, validation gates, worktrees, voice input, and optional orchestration tools. It maps directly to Sid's ai-memory and tech-radar development loop: preserve durable context, separate raw extraction from findings, promote repeated rules into memory or skills, and gate agent output with deterministic verification before merge or deployment.
What the post showed
Caption: Find everything from me here: https://linktr.ee/kunchenguid
Tools I mentioned:
- WezTerm https://wezterm.org/index.html
- tmux https://github.com/tmux/tmux/wiki
- Neovim https://neovim.io/
- npx skills CLI https://github.com/vercel-labs/skills
- OpenSuperWhisper https://github.com/starmel/OpenSuperWhisper
- AXI https://axi.md/
- lavish https://github.com/kunchenguid/lavish-axi
- no-mistakes https://github.com/kunchenguid/no-mistakes
- gnhf https://github.com/kunchenguid/gnhf
- treehouse https://github.com/kunchenguid/treehouse
- firstmate https://github.com/kunchenguid/firstmate
--------
Chapters: 00:00 How to watch this video 02:10 Why I work in the terminal 03:49 WezTerm and my lua config 05:09 What is tmux 07:18 Neovim as editor 09:21 Agent harnesses 11:24 My global memory file 14:50 Project level memory file 16:12 Using skills 18:40 How skills may hurt your agent 20:27 Voice input 22:12 The importance of agent ergonomics 24:25 Planning with interactive artifacts 28:29 Validating code changes 33:12 Long running tasks 36:20 Parallel worktrees and agents 40:23 First mate 44:26 The captain's mindset
Key claims from transcript: Hi everyone. Welcome to this video and this will be a full walkthrough of my agent engineering workflow. My name is Kun. I was previously an late principal engineer, worked at meta, Microsoft, and assassin on many large scale systems like the Bing search engine, windows, and Facebook games. In the recent couple of years, I have been building frontier coding agents at Atlassian and helped many engi
Learning chapters:
- 00:00 How to watch this video
- 02:10 Why I work in the terminal
- 03:49 WezTerm and my lua config
- 05:09 What is tmux
- 07:18 Neovim as editor
- 09:21 Agent harnesses
- 11:24 My global memory file
- 14:50 Project level memory file
- 16:12 Using skills
- 18:40 How skills may hurt your agent
- 20:27 Voice input
- 22:12 The importance of agent ergonomics
- 24:25 Planning with interactive artifacts
- 28:29 Validating code changes
- 33:12 Long running tasks
- 36:20 Parallel worktrees and agents
- 40:23 First mate
- 44:26 The captain's mindset
On-screen text / OCR: agentic aes : ° r, OD. F fie engineering “A \ Worktree ye. validation skills pie! | wy Tees review = clan parallel +mu% Ls a ‘. i : Mm workflow | hte WhISpEe 4
Extraction path:
- yt-dlp:metadata
- yt-dlp-cli:comments
- youtube-transcript-api
Source links found:
- https://linktr.ee/kunchenguid
- https://wezterm.org/index.html
- https://github.com/tmux/tmux/wiki
- https://neovim.io/
- https://github.com/vercel-labs/skills
- https://github.com/starmel/OpenSuperWhisper
- https://axi.md/
- https://github.com/kunchenguid/lavish-axi
- https://github.com/kunchenguid/no-mistakes
- https://github.com/kunchenguid/gnhf
- https://github.com/kunchenguid/treehouse
- https://github.com/kunchenguid/firstmate
Linked artifacts:
- profile · creator/profile link: https://linktr.ee/kunchenguid
- terminal_tool · terminal cockpit: https://wezterm.org/index.html
- github_repo · linked GitHub repository: https://github.com/tmux/tmux/wiki
- docs · documentation site: https://neovim.io/
- skill · agent skill system: https://github.com/vercel-labs/skills
- voice_input · voice input tool: https://github.com/starmel/OpenSuperWhisper
- agent_interface · agent-facing CLI/interface pattern: https://axi.md/
- interactive_planning · interactive planning artifact: https://github.com/kunchenguid/lavish-axi
- validation_gate · pre-push validation gate: https://github.com/kunchenguid/no-mistakes
- long_running_agent · long-running agent loop: https://github.com/kunchenguid/gnhf
- worktree_orchestration · parallel worktree management: https://github.com/kunchenguid/treehouse
- agent_orchestration · agent crew coordination: https://github.com/kunchenguid/firstmate
Top comments:
- 10 bounded YouTube comments were captured.
- Highest-signal viewer theme: people wanted to feed the workflow back into their own agents and evaluate which ideas to adopt.
- Comment evidence is treated as untrusted social context, not as installation guidance.
Workflow Audit
Workflow type: agentic engineering workflow
Recommended intake:
- Treat this as a workflow/harness input, not just a tool recommendation.
- Map stable rules to memory, reusable instructions to skills, and risky automation to explicit validation gates.
- Do not install the full stack by default; pilot the smallest artifact that improves the current harness.
Validation gates:
- pre-push validation gate: https://github.com/kunchenguid/no-mistakes
Planning artifacts:
- interactive planning artifact: https://github.com/kunchenguid/lavish-axi
Long-running loops:
- long-running agent loop: https://github.com/kunchenguid/gnhf
Orchestration candidates:
- parallel worktree management: https://github.com/kunchenguid/treehouse
- agent crew coordination: https://github.com/kunchenguid/firstmate
Ergonomics and interface tools:
- terminal cockpit: https://wezterm.org/index.html
- agent skill system: https://github.com/vercel-labs/skills
- voice input tool: https://github.com/starmel/OpenSuperWhisper
- agent-facing CLI/interface pattern: https://axi.md/
What it actually is
- What: Kun Chen's L8 Principal agentic engineering workflow is a terminal-centered AI development harness that combines memory files, reusable skills, interactive planning artifacts, validation gates, worktrees, voice input, and optional orchestration tools.
- Who built it / maintained by: Kun Chen
- Status: experimental
- How it compares to alternatives:
- Compared with a single chat session, it makes agent work resumable through global/project/session memory.
- Compared with ad hoc prompting, it turns recurring behavior into reviewed skills and validation gates.
- Compared with autonomous multi-agent orchestration first, it starts with evidence quality, ergonomics, and verification before adding crew-style agents.
- GitHub stars: 0 · License: unknown · Archived: no
Links
Kickstarter guide
Keep Sid's current Codex/Cursor flow as the cockpit; do not switch editors just to copy the workflow. Use the audit doc as the operating map: /docs/handoff/2026-07-05-kun-agentic-workflow-memory-audit.md. Promote only stable repeated rules to GLOBAL_MEMORY.md or project MEMORY.md. Pilot Lavish/AXI-style planning for one complex audit, and use no-mistakes as validation-gate reference architecture before adopting hooks globally. Process queued child artifacts as separate findings before installing tools.
Child Artifact Intake Status
- WezTerm:
2026-07-06-wezterm-wez.md - tmux:
2026-07-06-home.md - Neovim:
2026-07-06-neovim.md - Vercel Skills:
2026-07-06-github-vercel-labsskills-the-open-agent-skills-tool-npx-skil.md - OpenSuperWhisper:
2026-07-06-github-starmelopensuperwhisper-macos-dictation-app.md - AXI:
2026-07-06-axi-agent-experience-interface.md - lavish-axi:
2026-07-06-github-kunchenguidlavish-axi-html-is-the-new-markdown-lavish.md - no-mistakes:
2026-07-06-github-kunchenguidno-mistakes-git-push-no-mistakes.md - gnhf:
2026-07-06-github-kunchenguidgnhf-before-i-go-to-bed-i-tell-my-agents-g.md - treehouse:
2026-07-06-github-kunchenguidtreehouse-manage-worktrees-without-managin.md - firstmate:
2026-07-06-github-kunchenguidfirstmate-talk-to-one-agent-ship-with-a-cr.md