whoami — instant user context for AI coding agents

Stop re-explaining yourself to every AI agent. whoami scans your machine and interviews you once, then writes a portable profile any AI coding assistant can load in seconds.

What is whoami?

whoami is an open-source Claude Code skill that builds a complete, portable profile of who a developer is and what they do, so any AI coding agent can load instant context instead of starting cold. It combines an automated machine scan with a short onboarding interview, then writes a single Markdown profile (~/.whoami/profile.md) that LLM agents — Claude Code, Cursor, Codex, and others — can read in seconds.

In one line: whoami gives your AI agent memory of you before the conversation even starts.

Why developers need it

Every new AI agent session starts from zero. It doesn't know your name, your tech stack, the projects you're building, or how you prefer to work — so you re-explain your setup over and over, every session, with every tool. That repeated context-setting is wasted time and produces worse agent output.

whoami solves the AI agent onboarding problem permanently: capture your context once, reuse it everywhere.

Features

• 🧠 Machine context scanning — reads the developer signals already on your disk
• 🎤 Grounded onboarding interview — captures intent and goals your files can't reveal
• 🤖 Cross-agent session awareness — sees your history across Claude Code, Codex, Droid/Factory, OpenCode, Gemini, and Qwen
• 📄 Portable profile — one dense profile.md any LLM agent can load as context
• 🔒 Local & private by default — zero network calls, secrets redacted, your data never leaves the machine
• ⚡ One command — run /whoami in Claude Code, or bash scan.sh standalone
• 🪪 MIT licensed — free, open source, auditable

How it works

whoami works in two passes, combined into one profile:

1. Machine scan (scan.sh) — objective signals already on disk:
   • git identity & repositories (remotes, recent commits, dominant languages)
   • installed toolchains, languages, and global packages
   • editor & dotfile config (presence only)
   • AI-agent session history across CLIs — Claude Code, Codex, Droid/Factory, OpenCode, Gemini, Qwen (counts, recent project directories, recent prompts)
   • shell history (most-used + recent commands)
   • normal-file structure (folder layout, recently modified files)
2. Onboarding interview — the skill asks many grounded questions (made specific by what the scan found) to capture intent, priorities, working style, and goals the machine can't infer.

Output: ~/.whoami/profile.md (plus scan-raw.txt), optionally persisted into Claude Code memory so it applies automatically in future sessions.

Quickstart

Install (recommended — Claude Code plugin)

Add the marketplace and install in two commands inside Claude Code:

/plugin marketplace add luckeyfaraday/whoami
/plugin install whoami@luckeyfaraday

That's it — /whoami is now available. Update later with /plugin marketplace update luckeyfaraday.

Install (manual / for hacking on it)

Clone the repo and symlink the skill into your Claude Code skills directory:

git clone https://github.com/luckeyfaraday/whoami.git
ln -sfn "$PWD/whoami/skills/whoami" ~/.claude/skills/whoami

Use

In Claude Code, run the skill:

/whoami

…or just ask: "onboard me" / "figure out who I am". Re-running is idempotent — it diffs against your existing profile and only asks about what changed.

Run the scan standalone (no agent required):

bash scan.sh --max-repos 40 --depth 2

Flags

flag	default	meaning
--home DIR	$HOME	root directory to scan
--max-repos N	40	cap on git repositories reported
--depth N	2	folder-tree depth for normal files

What the profile looks like

# whoami — <name>
## Snapshot          # 2–3 sentence elevator description
## Identity
## Role & focus
## Current work      # active projects, priorities, goals (dated)
## Expertise & stack
## Tooling & environment
## Working style & preferences   # how agents should behave with you
## Constraints & boundaries
## Goals
## Open questions / low-confidence

Privacy

whoami reads your machine, so privacy is the core design constraint — not an afterthought:

• Conservative scope by default — dev/work signals + normal-file structure only.
• Secrets are redacted — lines matching key / token / password patterns are dropped.
• Personal-doc contents are never printed — only names and paths.
• Everything stays local — there are zero network calls. The generated profile and raw scan live under ~/.whoami/ and are gitignored.
• Broader scope is opt-in — browser history, SSH known_hosts, and document contents are only collected if you ask, and the skill confirms first.

scan.sh is a single, readable shell script — audit it before you run it.

Supported AI agents

whoami detects and reads session history from these AI coding assistants and CLIs:

Claude Code · OpenAI Codex · Droid (Factory) · OpenCode · Gemini CLI · Qwen Code — plus detection for Cursor, Continue, Aider, and Ollama.

The generated profile is plain Markdown, so it works as context for any LLM agent, including Cursor, Windsurf, GitHub Copilot, and custom agents built on the Claude or OpenAI APIs.

Use cases

• Onboard a new AI agent to your stack and projects in seconds
• Stop repeating context across Claude Code, Cursor, Codex, and other tools
• Standardize agent behavior — encode how you want agents to act (autonomous vs. ask-first, terse vs. explanatory)
• Bootstrap agent memory for a fresh machine or a new teammate
• Build a personal AGENTS.md / context file grounded in real signals, not guesses

FAQ

What is whoami? whoami is an open-source Claude Code skill that profiles a developer — identity, stack, projects, working style — by scanning their machine and interviewing them, producing a portable Markdown profile any AI agent can load as instant context.

Is whoami safe / private? Yes. It runs entirely locally with zero network calls, redacts secrets, never prints personal-document contents, and stores output only under ~/.whoami/ (gitignored). The scanner is a single readable bash script you can audit.

Does whoami work with agents other than Claude Code? Yes. While it installs as a Claude Code skill, the output is plain Markdown that works as context for any LLM agent — Cursor, Codex, Windsurf, GitHub Copilot, or custom agents on the Claude/OpenAI APIs. The scanner also runs standalone with bash scan.sh.

Where does whoami store its output? In ~/.whoami/ — profile.md (the profile) and scan-raw.txt (the raw scan). Both are gitignored and never leave your machine.

What data does whoami collect? Git identity and repos, installed toolchains and packages, dotfile presence, AI-agent session history, shell history (commands only), and normal-file structure. Secrets and personal-document contents are excluded by default.

How is whoami different from CLAUDE.md or AGENTS.md? Those are hand-written, static, and per-project. whoami auto-generates a user-level profile from real machine signals plus an interview, and keeps it current across every project and agent.

Contributing

Issues and pull requests are welcome. Good first additions: support for more agent CLIs, more package managers, and macOS/WSL coverage. Keep the privacy contract intact — no secret material, no personal-doc contents, and no network calls in scan.sh.

License

MIT © luckeyfaraday

---

_{Keywords: Claude Code skill · AI agent context · LLM context · developer profile · AI coding assistant onboarding · agent memory · machine context scanner · Cursor · Codex · personal context for AI · AGENTS.md alternative · CLAUDE.md generator}