Hi, I'm Kun, a member of the technical community. I previously worked as an L8 principal engineer at Meta, Microsoft and Atlassian.

This "/kun" skill here is a near-realtime distillation of my experience, knowledge, tools, workflows and skills. The instructions and knowledge base here is updated daily based on what I said and did.

Use this skill to apply my stack, think and build like a seasoned professional developer.

# install (global recommended)
$ npx skills add kunchenguid/kun -g

# in your agent
/kun how should I improve my AGENTS.md?
/kun fix this nasty bug!
/kun how would you build a product that...
/kun <literally anything>

This is also available as a Grok Bot.

The /kun skill file itself stays thin on purpose. On /kun, the agent runs the public pull script (no LLM), then reads from a local cache and follows it.

daily automation                    /kun question
      │                                   │
      ▼                                   ▼
┌──────────────────────┐         ┌──────────────────────┐
│ refresh living docs  │         │ node pull-kun.mjs    │
│ on main (see below)  │         │ → ~/.cache/kun       │
└──────────┬───────────┘         │ (or $KUN_PULL_DIR)   │
           │                     └──────────┬───────────┘
           ▼                                ▼
   OPINIONS.md  TOOLS.md              read ENTRY + docs
   VOICE.md  ENTRY.md                 (+ selective content/)
   content/ MANIFEST                        │
           \______________________________/ │
                          │
                          ▼
                   concrete answer

1. Download scripts/pull-kun.mjs from this repo (raw.githubusercontent.com, jsDelivr fallback) and run node … --dir <cache> (default: $KUN_PULL_DIR or ~/.cache/kun).
2. Read full local copies of ENTRY.md, TOOLS.md, OPINIONS.md, and VOICE.md from that cache.
3. When a question needs Kun's actual words, open matching files under content/ in the cache — do not dump the whole tree into context.
4. If the pull fails, stop and say so; do not guess.

No GitHub CLI and no GitHub auth are required for end users.

Automation runs in Grok Bot and updates this repo daily (America/Los_Angeles):

• OPINIONS.md and VOICE.md from Kun's public X, Substack, and YouTube. New signals are merged and tightened into the existing map first; append only when something is truly new.
• content/ archive of those public posts (plus content/MANIFEST.json) for raw-ledger / skill pulls via scripts/pull-kun.mjs.
• TOOLS.md for Kun-owned public, non-archived repos with a meaningful number of stars.

So /kun always reasons from the latest committed files on main, not from a frozen copy inside the skill package.

Raw public posts live under content/ (one markdown file per item) with content/MANIFEST.json for consumers. Keep OPINIONS.md / VOICE.md compact; do not treat content/ as a second opinions dump.

# Incremental sync into ~/.cache/kun (or $KUN_PULL_DIR / --dir)
node scripts/pull-kun.mjs

# Or download the script alone and pull into a cache:
#   https://raw.githubusercontent.com/kunchenguid/kun/main/scripts/pull-kun.mjs
#   (jsDelivr: https://cdn.jsdelivr.net/gh/kunchenguid/kun@main/scripts/pull-kun.mjs)
node /path/to/pull-kun.mjs --dir ~/.cache/kun

pull-kun.mjs fetches remote content/MANIFEST.json (raw.githubusercontent.com, jsDelivr fallback), downloads only new/changed content files, deletes local files removed from the manifest, and hash-syncs ENTRY.md, VOICE.md, OPINIONS.md, and TOOLS.md. First empty cache = full pull once; thereafter incremental.

This repo is literally Kun's own knowledge base so it deliberately does not accept PR contributions.

Bug reports and suggestions are welcome as issues!