Find agent skills by outcome

- For exploratory tasks

prefer action over asking unless information is required for later steps

- Give concrete facts

not abstract comparisons

- Share recommendation with reasoning

then ask user to decide

- If hypotheses disproven

generate new ones from gathered information

- Policies

rules

How: Key implementation decisions (patterns used

structure chosen)

- On transient errors

retry (unless explicit limit reached)

- When multiple valid approaches: speculate

think aloud

- Once action taken

it cannot be undone

- For each option: WHY it exists

HOW it works

Turkish Cats hanging out nearby of Galata Tower

Turkish Cats hanging out nearby of Galata Tower, vertical

If you can track internally

do it internally.

- Identify gaps

ambiguities

- Scan package.json

Cargo.toml

If not

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

Before any action

gather context and enhance the request internally:

Ultrathinker

# Ultrathinker

- Use soft

warm tones to evoke a familial and welcoming atmosphere.

Productivity

- Each task completable in a scoped

small change

2. Apply Deepthink Protocol (reason about dependencies

risks

- Policy rules

mandatory prerequisites

- Look for CLAUDE.md

AGENTS.md

- Check for .claude/ folder (agents

commands

2. Propose enhancements to tools

processes