Find agent skills by outcome

1. Accept parameters such as ${platform:YouTube/Database/Profile}

${metrics:engagement/views/likes}

- Once action taken

it cannot be undone

- If hypotheses disproven

generate new ones from gathered information

- Give concrete facts

not abstract comparisons

- Share recommendation with reasoning

then ask user to decide

- When multiple valid approaches: speculate

think aloud

- On transient errors

retry (unless explicit limit reached)

- For each option: WHY it exists

HOW it works

How: Key implementation decisions (patterns used

structure chosen)

- Policies

rules

2. Apply Deepthink Protocol (reason about dependencies

risks

If not

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

- Look for CLAUDE.md

AGENTS.md

- Policy rules

mandatory prerequisites

Ultrathinker

# Ultrathinker

Before any action

gather context and enhance the request internally:

- Check for .claude/ folder (agents

commands

- Scan package.json

Cargo.toml

- Identify gaps

ambiguities

If you can track internally

do it internally.

- Use soft

warm tones to evoke a familial and welcoming atmosphere.

Turkish Cats hanging out nearby of Galata Tower

Turkish Cats hanging out nearby of Galata Tower, vertical

1. Provide unit

integration

3. Deliver results in Markdown

JSON/YAML