THE MASKED FOUNDERANONYMOUS BUILD LOG
The toolbox . skills library

Teach your AI
how you work.

7 skills, pulled from how I actually run my companies. A skill is a protocol your AI loads and follows, every time. Each card shows what it does, when to use it, and the exact file to copy.

Start with The Handoff and The Lessons List. Those two change everything else.

Skill

The Handoff

Ends every session by saving state, so the next one starts warm instead of from zero.

When: At the end of any work session, and any time the AI's context is getting long.

How it works

  • Write DONE (with proof), IN FLIGHT (exactly where it stopped), and NEXT.
  • List everything waiting on the human, by name.
  • Next session opens by reading the handoff before touching anything.
Skill

The Grill

Stress-tests a plan with one question at a time BEFORE any building starts.

When: Before starting anything expensive: a feature, a purchase, a launch, a pivot.

How it works

  • The AI maps the decisions your plan depends on, in order.
  • It asks ONE question at a time, each with a recommended answer and why.
  • Nothing gets built until you both agree on the whole picture.
Skill

Verify Before Done

Blocks the word 'done' until there is proof for every part of the ask.

When: At the end of every task an AI did for you, before you trust or ship it.

How it works

  • Turn the original ask into a checklist.
  • Each item needs pointable proof: the output itself, not 'I did it'.
  • Check the REAL thing: the rendered page, the sent email, the actual file.
Skill

The Debug Protocol

Replaces guess-and-pray debugging with a fixed order: reproduce, isolate, prove, fix, prove again.

When: The moment anything behaves unexpectedly, before trying fixes.

How it works

  • Reproduce it on demand first. No reproduction, no fixing.
  • Isolate: cut the problem space in half until the cause is cornered.
  • Prove the cause, apply ONE fix, then prove the fix on the original case.
Skill

Ship Check

The pre-launch gate: clean state, real build, live verification. Catches the 'it worked on my machine' family.

When: Right before anything goes live: a site, a post, a product change.

How it works

  • Nothing uncommitted or half-finished rides along with the release.
  • The build/test command must show its own exit code, not a filtered log.
  • After launch, verify the LIVE surface in a real browser, not localhost.
Skill

The Lessons List

Turns every mistake into a one-line rule the AI reads on every future session. The list only grows.

When: Every time you correct the AI, or it catches its own mistake.

How it works

  • Same day, the general lesson gets appended as one line. Not the story, the rule.
  • Lessons are never deleted or softened.
  • The file loads into every session, so the same mistake costs you once.
Skill

The Daily Journal

Mines your own work history each day for the 3 to 5 things actually worth remembering or sharing.

When: End of each working day, or every morning over yesterday's work.

How it works

  • The AI reads what actually happened: commits, notes, decisions, mistakes.
  • It drafts the day's learnings, each traced to the real moment it came from.
  • Nothing invented: a quiet day produces a short entry or none at all.

A skill is not magic. It works because your AI reads it every time, so the good habit stops depending on your memory. Built for Claude Code, but the pattern works in any AI tool.

Go deeper

Skills shape the work. Agents do the work.

Pair these with the agent blueprints and the daily prompts. One vault, three tools.

Claude Code Skills: My Free Skills Library . The Masked Founder