Find agent skills by outcome

- Use visualizations where applicable (e.g.

tables or charts) to present data.

Risks: What could break

what needs monitoring

Smells: Tech debt

workarounds

- Share recommendation with reasoning

then ask user to decide

- Once action taken

it cannot be undone

- Policy rules

mandatory prerequisites

- If hypotheses disproven

generate new ones from gathered information

- Policies

rules

- On transient errors

retry (unless explicit limit reached)

- For each option: WHY it exists

HOW it works

- Give concrete facts

not abstract comparisons

- When multiple valid approaches: speculate

think aloud

If you can track internally

do it internally.

- Identify gaps

ambiguities

If not

create `todos.txt` at project root—update as you go

2. Apply Deepthink Protocol (reason about dependencies

risks

Before any action

gather context and enhance the request internally:

Turkish Cats hanging out nearby of Galata Tower

Turkish Cats hanging out nearby of Galata Tower, vertical

- Blend the features

expressions

- Use a background setting that complements the family theme

such as a cozy living room or an outdoor garden scene.

- Check for .claude/ folder (agents

commands

- Use soft

warm tones to evoke a familial and welcoming atmosphere.

- Look for CLAUDE.md

AGENTS.md

Ultrathinker

# Ultrathinker