Prism

Thinking lenses.

Prism is a skill for Claude Code that looks at your task through a few lenses first, so you get a thought-out answer instead of a quick, generic one.

Usage

• do X for me — direct, as usual.
• /prism do X for me — it picks the lenses that fit and lays out the decisions first.

Install

It's a plain folder-skill.

Let Claude do it — paste to Claude Code:

Install the skill from https://github.com/xtompie/prism into ~/.claude/skills/prism

Or yourself:

git clone https://github.com/xtompie/prism.git ~/.claude/skills/prism

Restart the session and /prism is available. Update with git pull in that folder. It has disable-model-invocation: true, so it never runs on its own — only when you type /prism.

What's a lens?

A reusable way of looking at a task — not a solution, not domain knowledge, not a procedure. It doesn't give the answer; it changes the angle and raises a question you'd otherwise skip.

inversion — instead of asking "how do I make this succeed?", ask "how would I make it fail?" For a team trip, "how do I make it great?" gives vague answers. "How would I ruin it?" gives clear ones — no budget, too few rooms, booked too late, no plan for rain. Then turn each one into a must-do.

Where lenses come from

Prism looks in three places:

• The model's own head — lenses it already knows.
• The notes — a folder of short lenses that grows over time.
• The web — only when you ask — to find a new way of looking for the task.

When a new lens proves useful, Prism offers to save it to the notes, with a check against duplicates. The base grows from use.

What's in the repo

• SKILL.md — the process: pick lenses, lay out decisions, grow the base. No domain knowledge.
• INDEX.md — the lens menu: name + when to use.
• lenses/*.md — each lens in five lines: name / when / when_not + the move.

The model reads the INDEX and loads only the few lenses that fit a task, not the whole set.