Find agent skills by outcome

Productivity

- For exploratory tasks

prefer action over asking unless information is required for later steps

- When multiple valid approaches: speculate

think aloud

- Share recommendation with reasoning

then ask user to decide

- Once action taken

it cannot be undone

- For each option: WHY it exists

HOW it works

- On transient errors

retry (unless explicit limit reached)

- If hypotheses disproven

generate new ones from gathered information

How: Key implementation decisions (patterns used

structure chosen)

- Give concrete facts

not abstract comparisons

- Policies

rules

- Identify gaps

ambiguities

Ultrathinker

# Ultrathinker

If you can track internally

do it internally.

- Look for CLAUDE.md

AGENTS.md

Productivity

- Each task completable in a scoped

small change

- Use soft

warm tones to evoke a familial and welcoming atmosphere.

- Scan package.json

Cargo.toml

- Check for .claude/ folder (agents

commands

If not

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

PromptBeginner5 minmarkdownQuality: 24

Turkish Cats hanging out nearby of Galata Tower

Turkish Cats hanging out nearby of Galata Tower, vertical

2. Apply Deepthink Protocol (reason about dependencies

risks

Before any action

gather context and enhance the request internally:

- Policy rules

mandatory prerequisites