onboard
Skill1 starsChanged 4 months agoMITView on GitHub
Guided interview that builds your personal context system — the foundation every digital employee reads from. Fills profile.md, company.md, and decision-style.md by interviewing you section by section, only asking for what's missing. Use when you say "onboard", "set me up", "fill my context", "get started", or run this first before any other skill.
168 other projects use a version of this.
Adopted by
Most recently taken up by
- jestatsio/codebase-mentortheir own version, first seen Nov 15, 2026
- chuanglibs/Claude-Code-Game-Studios-Codextheir own version, first seen Nov 15, 2026
- 5uck1ess/devkittheir own version, first seen Nov 15, 2026
- suitedaces/dorabottheir own version, first seen Nov 14, 2026
- danielrosehill/Claude-Audio-Production-Plugintheir own version, first seen Nov 13, 2026
- Donchitos/Claude-Code-Game-Studiostheir own version, first seen Nov 9, 2026
Used with
Projects with it also use soak-test, team-live-ops and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
lilangMaxSkillsoak-test
Generate a soak test protocol for extended play sessions. Defines what to observe, measure, and log during long play sessions to surface slow leaks, fatigue effects, and edge cases that only appear after sustained play. Primarily used in Polish and Release phases.
Also in 9 projects that use this
lilangMaxSkillteam-live-ops
Orchestrate the live-ops team for post-launch content planning: coordinates live-ops-designer, economy-designer, analytics-engineer, community-manager, writer, and narrative-director to design and plan a season, event, or live content update.
Also in 9 projects that use this
HK-hubSkillteach-impeccable
One-time setup that gathers design context for your project and saves it to your AI config file. Run once to establish persistent design guidelines.
Also in 19 projects that use this
CoralGameSkillproject-stage-detect
Automatically analyze project state, detect stage, identify gaps, and recommend next steps based on existing artifacts.
Also in 15 projects that use this
CoralGameSkillteam-level
Orchestrate level design team: level-designer + narrative-director + world-builder + art-director + systems-designer + qa-tester for complete area/level creation.
Also in 11 projects that use this
- +5 more from lilangMax
- +2 more from CoralGame
Other versions168 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 168
- Donchitos/Claude-Code-Game-Studios★ 39ksame name
- parcadei/Continuous-Claude★ 3.9ksame name
- parcadei/Continuous-Claude-v2★ 3.9ksame name
- parcadei/Continuous-Claude-v3★ 3.9ksame name
- nateherkai/AIS-OS★ 1.4ksame name
- poshan0126/dotclaude★ 871same name
- talesofai/cohub★ 585same name
- me2resh/apexyard★ 542same name
- nexmoe/DayBee★ 424same name
- nexmoe/eve★ 424same name
158 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository12 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- brain-dumpSkill
- chief-of-staffSkill
- comms-expertSkill
- docx-reviewSkill
- handoffSkill
- interview-meSkill
- ops-powerhouseSkill
- research-analystSkill
- sort-knowledgeSkill
- strategic-advisorSkill
- wiki-lintSkill
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 6 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 156 edited versions we read keep these.
6
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 36
Before asking anything, detect what already exists so you never re-ask:
- Very specific: line 38
Check for each of context/profile.md, context/company.md, context/decision-style.md.
- Distinctive: line 52
If all three exist and are current, confirm there's nothing to do and point the user at next steps (Section 7 / Step 4) instead of …
- Distinctive: line 101
Use AskUserQuestion here. Suggested prompts/options:
- Distinctive: line 112
Follow up conversationally: "Which topics do you want MORE pushback on?
- Distinctive: line 135
Before writing files, show the user a tight summary of what you're about to save for each file and get a quick confirmation (Operating …
- Distinctive: line 149
Write the three files (overwriting only the sections that changed if a file already existed.
“Common” is measured: kept by at least 6 in 10 of the 156 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/onboard — build the foundation everything reads from
1. Purpose
This is the first thing the user runs. The lecturer is emphatic: "I strongly recommend that you build a personal context system... it feeds into every strategic conversation. And that's the foundation. Everything else is built upon." Without it, every other employee gives generic output; with it, "you get something that feels like an advisor who've been working for you for years." Your job is to be the interviewer who extracts that context out of the user's head — their role, mandate, company, market, competitive stance, and the way they like to decide — and persist it so they never re-explain themselves.
You are not filling in a form. You are conducting a warm, sharp onboarding interview, like a great chief of staff on their first week getting to know a new principal. Pull what already exists, ask only for the gaps, and write it down.
2. Operating principles applied
See .claude/OPERATING_PRINCIPLES.md. This skill is the purest expression of:
- #3 Have AI interview you — this entire skill IS the interview. Grill the user section by section; surface gaps they wouldn't have volunteered.
- #1 Speak, don't type — explicitly invite dictation. The user's messy, out-loud thinking is the most valuable input; nothing here penalizes it.
- #5 Intentional intervention — this is the up-front primer capture for the whole system. Everything downstream inherits the judgment captured here.
3. Step 1 — Load context
Before asking anything, detect what already exists so you never re-ask:
- Check for each of
context/profile.md,context/company.md,context/decision-style.md. - For each file that exists, read it and summarize back to the user what's already known: "Here's what I already have on file for you..." Then ask only whether anything is stale and needs updating — skip sections that are complete and current.
- For each file that does NOT exist, read the matching
*.example.mdtemplate (context/profile.example.md,context/company.example.md,context/decision-style.example.md) so your interview mirrors that exact structure, and plan to interview for it fully. - Tell the user the plan up front: "You have profile.md already — I'll just check it's current. company.md and decision-style.md are empty, so I'll interview you for those. About 10 minutes. You can dictate every answer."
If all three exist and are current, confirm there's nothing to do and point the user at next steps (Section 7 / Step 4) instead of re-interviewing.
4. Step 2 — Interview the user
Open by setting the tone: "I'm going to ask you a handful of questions to build your context system. Please talk, don't type — ramble, go on tangents, give me the messy version. I'll structure it. Hit me with voice notes if that's easier."
Go section by section. One focused question at a time for open-ended fields; AskUserQuestion for the structured choices. Don't dump all questions at once. Confirm each section back in your own words before moving on, and keep probing when an answer is thin ("What else? What would your team say you're actually measured on?").
A. profile.md — who the user is
Conversational questions (ask, listen, follow up):
- "What's your title, and what are you actually accountable for — the 2–3 outcomes you're measured on, not the job description?"
- "What's your scope? Team size, budget, regions, P&L ownership?"
- "Where is your judgment strongest — where should AI defer to you? And where are you newer or want to be pushed harder?"
- "Your top 3 priorities this quarter?"
- "What decisions do you make over and over?" (hiring, pricing, roadmap, deals)
- "How do you work? Time zone, when you do deep work, how you like to consume a briefing. And the pet peeves — what should never show up in output to you?"
- "What does 'good' look like to you? The quality bar you hold."
B. company.md — company, market, competitive stance, sources
Conversational questions:
- "What does the company do, in one line? Stage, size, how you make money?"
- "What category do you play in, and who exactly do you sell to — your ICP?"
- "Where do you think the market is heading?"
- "Main competitors, and how you're different? Your moat? Where you're vulnerable?"
Then the sources field — the lecturer is explicit that research must be personalized to trusted sources and that some sources should be excluded. Ask this carefully because the Research Analyst depends on it:
- "What are your reliable internal data sources — P&L, metrics, CRM, dashboards?"
- "What external sources do you trust, in priority order?"
- "Which sources should I never treat as authoritative?" (e.g. vendor marketing, a competitor's blog, low-signal news). Persist this as an explicit EXCLUDE list — it's as important as the trusted list.
- "Anything big that happened last quarter that AI should know about?"
C. decision-style.md — how the user likes to decide
Use AskUserQuestion here. Suggested prompts/options:
- Default decision mode (single-select):
- Enumerated options — "Give me 5–10 clear options to pick from"
- Bottom line first — "One recommendation after the debate, then the why"
- Push hard, then space — "Challenge me hard, then let me sit with it"
- Socratic — "Ask me questions until I decide myself"
- Pushback calibration (single-select scale):
- 1–3 — Just execute / 4–6 — Balanced challenge / 7–8 — Push hard / 9–10 — Fight me on everything
- Follow up conversationally: "Which topics do you want MORE pushback on? And which have you already decided, so I should just execute?"
- Inputs before deciding (multi-select, allow "Other"):
- Always show the downside case / Always flag reversible vs. irreversible / Always show what it costs to be wrong / Always name who's affected
- Pace — conversational: "What do you decide fast, and what do you deliberately sleep on?"
Throughout: if the user already answered something in an existing file, don't
ask — confirm. If an answer reveals something reusable beyond these three files
(a key stakeholder, a strong opinion), note it and offer to capture it later via
/brain-dump.
5. Step 3 — Do the work
Synthesize the interview into clean, structured markdown that mirrors the
exact section structure of each *.example.md template — same headings, same
order — but populated with the user's real words. Write in the user's voice
where it's a quote or opinion; tighten rambling dictation into crisp bullets
without sanitizing the personality out of it.
Before writing files, show the user a tight summary of what you're about to save for each file and get a quick confirmation (Operating Principle #4 — separate the plan from the execution). Let them correct anything.
6. Step 4 — Output & persist
Wiki Contract (
.claude/WIKI.md)./onboardbootstraps the wiki. After writing the three pages, initialize the control files if they don't exist: createcontext/index.mdfromcontext/index.example.md(with entries for the pages you just wrote,[[profile]]·[[company]]·[[decision-style]], dated) andcontext/log.mdfromcontext/log.example.md. Cross-link the pages where natural. Append anONBOARD:line tocontext/log.md(date "+%F %H:%M").
Write the three files (overwriting only the sections that changed if a file already existed; otherwise create fresh):
context/profile.mdcontext/company.mdcontext/decision-style.md
Then tell the user exactly what you saved and where, e.g.:
Saved:
context/profile.md— role, mandate, priorities, quality barcontext/company.md— what you do, ICP, competitive stance, trusted + excluded sourcescontext/decision-style.md— bottom-line-first mode, pushback at 7/10These are git-ignored, so your real data stays on your machine.
Finally, hand off to the rest of the system:
- Writing samples: "Drop 2–3 samples of your best writing into
context/writing-samples/(emails, memos, posts — by document type). When you run/comms-expertit'll analyze your voice from them." - Brain-dump habit: "From here on, after any meeting with undercurrents only
you can read, run
/brain-dumpand just talk. That undocumented context is what makes every other employee non-generic." - What to run next: point them to the digital employees (e.g. research analyst, strategic advisor, comms expert, ops powerhouse) now that the foundation exists.
7. Pro tips
- Don't accept thin answers. The whole value is in the unknown-unknowns. If someone says "I'm accountable for growth," push: "Growth of what, measured how, by when, versus what baseline?" That specificity is what makes the system smart.
- The EXCLUDE source list is a power move. Most people only think about trusted sources. Forcing the question "what should I never treat as authoritative?" prevents the Research Analyst from laundering garbage into a brief. Always get it.
- Capture opinions, not just facts. "Where the market is heading (your view)" and "where we're vulnerable" are where the user's judgment lives — that's the primer (Principle #5). Generic facts AI can fetch; the user's read it cannot.
- Onboarding is never finished. Tell the user this is a living system. As
priorities shift, re-run
/onboardand it'll only touch what's stale. - Keep it to ~10 minutes. Senior people bail on long forms. Move briskly, confirm in your own words, and let dictation do the heavy lifting.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

