8 Comments
User's avatar
Marketlab's avatar

This is interesting and it's very useful when you don't have any ideas for your notes it's really hard to write consistently without running out of ideas

Fatolu Adedoyin Juliana's avatar

Thanks for sharing 🤍

Albert Blibo's avatar

I’ll give it a go. Thank you for sharing.

M J Tringle's avatar

Good stuff. thank you for sharing !

AI Skill of the Week's avatar

Thank you! Glad you're finding it useful : )

John Hardt's avatar

SOP: Using Your Uploaded Skills

Skill Trigger it by saying What it does

Honest Thinking Partner "Be honest with me," "stress-test this," "what am I missing," or just paste a plan/decision No-praise breakdown: surfaces hidden assumptions, names the flawed logic, calculates the cost of what you're avoiding, ends with a concrete action plan + kill switch

humanAIzer "Humanize this," "de-AI this," "make this sound less robotic" + paste the text Strips AI vocabulary/structural tells (rule of three, "it's not X it's Y," flat rhythm) and rewrites with real voice. Paste a writing sample first if you want it matched to your voice specifically

Substack Notes "Write a Substack note," "turn this into a note" Picks the right note formula (contrarian, pattern observation, build-in-public, etc.) from 10 proven types and writes it in your voice, formatted for Substack — requires you to upload substack-notes.zip via Settings → Capabilities first

All three fire automatically on the trigger phrases — you don't need to name them by title.

write 5 substack notes about the SOP I just created

Architected five distinct Substack notes showcasing AI skill workflow documentation

Architected five distinct Substack notes showcasing AI skill workflow documentation

1 — Pattern Observation

A pattern I've noticed with people who build custom AI skills: they build three, then forget which one does what.

So they retype the same instructions every session instead of just saying the trigger phrase.

Fixed it with one page. Skill name, trigger phrase, what it does. Three lines.

Now the tool remembers instead of me.

2 — Contrarian Statement

More AI tools won't fix your workflow.

A clear map of the tools you already have will.

I had three custom skills sitting unused because I couldn't remember what activated them. One SOP fixed that faster than adding a fourth skill ever would have.

3 — Build-in-Public Update

Built a one-page SOP this week for the custom skills I've uploaded to Claude.

Each one used to need a paragraph of context, retyped from scratch every time I wanted it. Now it's one trigger phrase per skill.

Setup time went from a wall of text to a sentence.

4 — List-Based Tactical

How to actually use the custom AI skills you've uploaded, instead of forgetting they exist:

Write down each skill's name and its one job

Note the exact phrase that triggers it

Keep it to one line per skill

Put it somewhere you'll see it before your next long prompt

Three skills, three lines, done.

5 — Direct Advice

Stop writing a paragraph of instructions every time you want a specific kind of output from a tool you already built.

Write the SOP once. Skill name, trigger phrase, what it does. Let the trigger carry the weight from now on.

Skip this step and you'll be re-explaining the same thing to the same tool six months from now.

John Hardt's avatar

From Claude - This file has a formatting issue worth fixing before you upload it: it contains two separate YAML frontmatter blocks stacked on top of each other (one for substack-note-generator, then another for substack-notes), which isn't valid — a skill file can only have one frontmatter block. I merged them, keeping the fuller name and description, and packaged it as a proper skill folder.

Sunita Kumar | 🎙️Clear Signal's avatar

I am glad I found your account early so I can catch up on the prior weeks.