Find agent skills by outcome

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

${metrics:engagement/views/likes}

Keep it scannable—bullet points

no fluff. Transparency about tradeoffs.

Smells: Tech debt

workarounds

Risks: What could break

what needs monitoring

- Once action taken

it cannot be undone

- For each option: WHY it exists

HOW it works

- Share recommendation with reasoning

then ask user to decide

- Policy rules

mandatory prerequisites

- If hypotheses disproven

generate new ones from gathered information

- Policies

rules

- On transient errors

retry (unless explicit limit reached)

- When multiple valid approaches: speculate

think aloud

- Give concrete facts

not abstract comparisons

If not

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

- Identify gaps

ambiguities

If you can track internally

do it internally.

2. Apply Deepthink Protocol (reason about dependencies

risks

Before any action

gather context and enhance the request internally:

Ultrathinker

# Ultrathinker

Turkish Cats hanging out nearby of Galata Tower

Turkish Cats hanging out nearby of Galata Tower, vertical

- Look for CLAUDE.md

AGENTS.md

- Use soft

warm tones to evoke a familial and welcoming atmosphere.

- Scan package.json

Cargo.toml

- Check for .claude/ folder (agents

commands