Find agent skills by outcome

Risks: What could break

what needs monitoring

Keep it scannable—bullet points

no fluff. Transparency about tradeoffs.

- Once action taken

it cannot be undone

How: Key implementation decisions (patterns used

structure chosen)

Education

Comments Explain WHY

Not WHAT:

- On transient errors

retry (unless explicit limit reached)

- When multiple valid approaches: speculate

think aloud

- Share recommendation with reasoning

then ask user to decide

- For each option: WHY it exists

HOW it works

- Give concrete facts

not abstract comparisons

Business

- On other errors

change strategy—don't repeat failed approaches

Productivity

- For exploratory tasks

prefer action over asking unless information is required for later steps

- Policies

rules

- If hypotheses disproven

generate new ones from gathered information

2. Apply Deepthink Protocol (reason about dependencies

risks

- Policy rules

mandatory prerequisites

Productivity

- Each task completable in a scoped

small change

- Identify gaps

ambiguities

If you can track internally

do it internally.

If not

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

- Scan package.json

Cargo.toml

- Look for CLAUDE.md

AGENTS.md

- Check for .claude/ folder (agents

commands

Before any action

gather context and enhance the request internally: