Find agent skills by outcome

- For exploratory tasks

prefer action over asking unless information is required for later steps

- If hypotheses disproven

generate new ones from gathered information

- On transient errors

retry (unless explicit limit reached)

- When multiple valid approaches: speculate

think aloud

If you can track internally

do it internally.

Productivity

- Each task completable in a scoped

small change

2. Apply Deepthink Protocol (reason about dependencies

risks

- Policy rules

mandatory prerequisites

- Identify gaps

ambiguities

- Check for .claude/ folder (agents

commands

- Look for CLAUDE.md

AGENTS.md

- Scan package.json

Cargo.toml

If not

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

Ultrathinker

# Ultrathinker

Before any action

gather context and enhance the request internally:

- Blend the features

expressions

- Use soft

warm tones to evoke a familial and welcoming atmosphere.

- Use a background setting that complements the family theme

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

Turkish Cats hanging out nearby of Galata Tower

Turkish Cats hanging out nearby of Galata Tower, vertical

3. Deliver results in Markdown

JSON/YAML

2. Propose enhancements to tools

processes

- Provide access to a variety of games

from retro to modern.

- Prioritize bugs based on severity

user impact

1. Provide unit

integration