BUILTBYBAS
Included with Praxis Academy membership

FLIGHT

An AI building coach for anyone building with AI, it asks the four questions, holds a gate before it builds, and never stops coaching while it does.

Your words are the programming language, and Pre-Flight is the coach. YOU bring the ideas or brainstorm with Pre-Flight, from your first feature to the ones you ship, Pre-Flight guides your learning. No server, no dependencies, no extra API key.

Begin descent

Pre-Flight Coaching is for everyone. It's for the vibe-coder who wants to understand their code. It's for the new dev who wants to learn the why behind what AI is building. I built the Pre-Flight Coaching system because in a world where anyone can build with AI, everyone should have the opportunity to understand what they are building.

Bas — BuiltByBas

It will not build until the checklist is complete.

A pilot never rolls onto the runway on a hunch. Pre-Flight runs the same four questions on every feature, and refuses to write code until each one is answered. That gate is the whole point.

DONE

What does the finish line look like?

The very last thing the user sees when it worked. Concrete, observable, no hand-waving.

ARM
CHECK

How will YOU verify it works?

Proof you can see yourself, never just trusting the machine's say-so that it worked.

ARM
OUT OF SCOPE

What are we deliberately not building?

The edges of this feature, named out loud, so scope cannot quietly creep mid-build.

ARM
CONTEXT

What can the build not see on its own?

The facts living only in your head, the thing the code has no way of knowing until you say it.

ARM
⊘ Until all four read ARMED, Pre-Flight holds short. Answer them and it says: “We are ready to build. Are you ready to build it?”

Want more than a one-line check? The optional Pre-Flight Secops check (/secops) turns that second question into a real test plan before you build, the cases worth checking and what each one proves, coached out, never handed over as a list. The single check still gates; this only deepens it.

See it hold the line

Four ways people try to skip the work. Click one and watch Pre-Flight name the rule and hold, never a gotcha, always a lesson.

pre-flight: holding the gate

It builds with you, then coaches the test.

Once the gate clears, Pre-Flight builds the feature alongside you. Then it does the part most tools skip: it walks you through testing the thing with your own eyes, on your own device, before anyone calls it done.

Then it loops.

Feature shipped, it points you straight at the next one and runs the four questions again. The discipline compounds. By the third feature, the checklist is muscle memory.

Build alongsideCoach the testNext feature

Arrive with one feature, or ten. Line them up and fly the run.

The default is one feature, start to finish. Show up with a whole session in mind and two optional modes let you line the work up and build it back to back, without ever loosening the gate.

/PLAN

Flight Plan

Name the features, put them in order, and surface what depends on what. A leaderboard needs scores saved first, so score-saving leads and the leaderboard sits behind it. You own which features matter; Pre-Flight sequences them and flags the dependencies.

/CHART

Chart a Course

Every feature clears its own comprehension checkpoint first, no exceptions. Only then does Pre-Flight build the queue back to back, then walks you through testing each one against its own CHECK, in order.

You don't even have to choose. Name several features at once and Pre-Flight routes you automatically: two or three into Flight Plan, four or more into Chart a Course, always announced, always with an off-ramp. And it holds a hard line, one Flight Plan and one Chart a Course at a time, never more.

⊘ Batching changes the cadence, never the gates. Every feature still passes its own checkpoint and is still tested with your own eyes. Default off, available any time: /plan, /chart, or just ask.

Drive it with commands

Native slash commands on Claude Code, typed as text on every other surface. Each one acts, or shows you its state.

/planMap and order your features and their dependencies, or show the current plan.
/chartGate, build, and test a feature queue back to back, or show the queue and per-feature status.
/secopsRun a Secops check: a verification test plan before building, plus a security sweep against a catalog of the blind spots AI code review misses, one category at a time.

Sharper where you've earned it. Full support where you haven't.

PILOT PROFICIENCY
0Calibrated to you

Pre-Flight reads how you answer across features and remembers it in a DECISIONS.md in your own project folder. Where you consistently nail a question, it loosens the scaffolding and trusts you. Where you're still finding your footing, it keeps full support in place.

It nudges your next rep.

At the close of each feature it names the single highest-leverage place to practice next, grounded in your actual work, not a generic tip. It works fine without the log, and noticeably better with it.

It coaches the person, not a script.

One quick question at the start sets your opening register. From there Pre-Flight adapts to your skill, your growth, and how you show up in the moment, and it never announces a word of it.

PATH 01

New to building

Plain language, no jargon, an everyday example before a technical one. The most support, offered before you ask.

Register: gentle, scaffold-first
PATH 02

Some scars

Uses the words you've earned, sharper questions, trusts you to reach first. Help arrives when you get stuck.

Register: pain-prevention
PATH 03

Built, never shipped

Light on building talk, you have it. Leans hard into the shipping gaps: verification, scope, "done" for a stranger.

Register: shipping discipline

Then it grows with you. Reading your decision log, it lightens up on the questions you've earned and keeps full support on the ones you still hedge, per question, dialing down as readily as up. No badges, no "you leveled up", you just feel a coach who knows you.

When you sayWhat Pre-Flight does
"I think it'll be fine"▸Reflects the hedge, asks for the check you'd actually run.
"Claude will figure it out"▸Names what the build can't see, asks what only you know.
"while we're at it…"▸Marks it a new task, protects the scope of this one.
"tests pass, we're good"▸Asks what a failing result would have looked like.
✦ The floor never moves: empathy is universal, scaffold when stuck. Your path sets where you start, never a ceiling on the help you can get.

Type /learn and learn the principle from your own code.

Not a Wikipedia article on the domain. Pre-Flight teaches the why using the exact thing you just built, then offers to go deeper. It also opens a “Ready to learn?” door at the end of every feature, there when you want it, a “not now” away when you don't.

A folder of markdown. Each file does one job well.

No build, no runtime, nothing to install. Interpretable context methodology: every file has a single clear purpose, so a stranger can read the whole coach and understand exactly how it thinks.

▸pre-flight-coach/The coach. Drop it in, Claude becomes Pre-Flight.
▸.claude/skills/Where Claude Code registers /learn, /plan, /chart, /secops.
▸learn/The /learn command folder.
▸SKILL.mdA thin door routing into learning-mode.md.
▸plan/The /plan command folder.
▸SKILL.mdDefines Flight Plan mode in full.
▸chart/The /chart command folder.
▸SKILL.mdDefines Chart a Course mode in full.
▸secops/The /secops command folder.
▸SKILL.mdDefines Pre-Flight Secops.
▸CLAUDE.mdActivates the persona and points at the rest.
▸identity.mdWho Pre-Flight is, who it serves, what it refuses.
▸rules.mdThe loop: intake, the four questions, the switch.
▸examples/The calibration tape: an index and fifteen worked dialogues.
▸index.mdThe map: one line per dialogue and when to load it.
▸01-happy-path.mdThe led arc, start to checkpoint.
▸02-gate-holding.mdThe gate named and held.
▸03-authority-split.mdVision theirs, craft decided and taught.
▸04-rough-day.mdEmpathy without losing the line.
▸05-prepared-user.mdA clear user, confirmed fast.
▸06-full-loop.mdThe whole arc, start to finish.
▸07-learning-mode.md/learn deferred, then grounded.
▸08-leveling.mdThe register shifts, never announced.
▸09-naming-the-rule.mdGates enforced out loud.
▸10-session-modes.mdBatched cadence, per-feature gates.
▸11-secops.mdThe CHECK deepened into a test plan.
▸12-pasted-spec.mdA spec is an idea, not an order.
▸13-path4-intake.mdA builder arrives; same gates.
▸14-path4-four-questions.mdThe intention compiler at Path 4.
▸15-ai-trust-hedge.mdThe machine grading its own homework.
▸reference/The frameworks: two folders and 18 files.
▸skills/Readable reference copies of the four command skills.
▸learn/The command folder.
▸SKILL.mdA thin door routing into learning-mode.md.
▸plan/The /plan command folder.
▸SKILL.mdDefines Flight Plan mode in full.
▸chart/The /chart command folder.
▸SKILL.mdDefines Chart a Course mode in full.
▸secops/The /secops command folder.
▸SKILL.mdDefines Pre-Flight Secops.
▸showcase/This page.
▸index.htmlThe interactive walkthrough you're reading now.
▸leading-the-idea.mdLeading a user to understand their own idea.
▸the-four-questions.mdThe canonical pre-flight checklist.
▸ai-collaboration.mdThe intention curriculum for building with AI.
▸levels.mdCalibrating register to the intake answer.
▸leveling.mdReading the decision log to meet you per question.
▸project-types.mdAdapting CHECK, CONTEXT, OUT OF SCOPE to the platform.
▸habits.mdThe three craft habits: git, logs, commit-before-risky.
▸build-mode.mdThe switch into building, coaching the test, the loop.
▸development-map.mdThe seven development areas and the debrief.
▸learning-mode.mdThe opt-in /learn mode and the offer.
▸session-modes.mdFlight Plan and Chart a Course, the two opt-in modes.
▸secops.mdThe opt-in pre-build Secops check (/secops).
▸security-blindspots.mdThe Secops catalog of AI-missed security blind spots.
▸principles-canon.mdEach skill mapped to its recognized source.
▸project-memory.mdRemembering a project across sessions.
▸failure-modes.mdThe six week-one anti-patterns.
▸inquiry-patterns.mdSocratic question templates by moment.
▸avoidance-tells.mdFive language patterns that signal dodging.

Cleared on every surface.

FULLEST

Claude Code

Pre-Flight reads and writes files directly. Your project memory, the DECISIONS.md that powers its leveling, and the build itself all live in your folder and carry across sessions automatically.

SUPPORTED

Claude.ai Projects

Connect the repo, pick the folder, say hello. Your first message brings Pre-Flight to life, even just “Hi.”

SUPPORTED

ChatGPT

Add the folder to a Project, or attach the files, and say hello. Pre-Flight runs on the interpretable context, its commands typed as text.

SUPPORTED

Codex

Open your repo and point Codex at the folder. The same markdown drives the coaching, its commands typed as text.

OPUSSONNETHAIKUany current Claude model
GPT-5o-SERIESCODEXany current frontier model
print("hello")———▶ coach ▶ build ▶ test ▶ repeat ———▶ Full Stack Pro

You arrive with Ideas.
You leave with something built, tested, and yours.

Pre-Flight meets you at your very first feature and stays with you through the ones after, it remembers how you answer, grows with you, and teaches the why from your own code. It comes with your Praxis Academy membership, along with the other ICMs I build and the content I create on the platform. Become a member, say hello, and start your climb.