Skip to content
Back to skills

Research

ASecurity

Research before designing or deciding. Read current official docs, study how established projects and peers solve the same problem, cross-check sources, and verify claims against the code. Use before choosing an approach, library, architecture, or vendor, when asked how others do it or for best practices, and when a claim, fact, or API detail is uncertain.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 22, 2026
ai-agentsgoapi

Works with

  • api

Security analysis

A100/100

Pro scans all 4 files and shows the line behind each finding

Scanned October 22, 2026

npx -y skills add skrrt-sh/skills --skill research --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Research?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Research
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/skrrt-sh-research/badge)](https://www.skillsdirectory.com/skills/skrrt-sh-research)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: research
description: Research before designing or deciding. Read current official docs, study how established projects and peers solve the same problem, cross-check sources, and verify claims against the code. Use before choosing an approach, library, architecture, or vendor, when asked how others do it or for best practices, and when a claim, fact, or API detail is uncertain.
---

# Research

Decide from evidence, not memory. This applies to code and to copy, sales, legal, and UX choices.

## Before deciding

- Read the current official docs for every library, API, and platform involved. Your memory may
  be outdated.
- Find how established projects and direct competitors solve the same problem. Study them; never
  copy. Pick the best approach and improve on it.
- Prefer a mature library over writing your own. A new service, framework, or runtime needs the
  user's approval.
- Use the latest stable versions and current idioms.

## Sources

- Cross-check every important fact with a second independent source.
- Rank sources: official docs, source code, standards bodies, professional associations, and
  government or legal publications; then maintainers and established projects; then practitioner
  discussions such as Reddit or forums. Vendor templates and marketing rank last.
- Read each source fully before using it.
- Prefer sources from the last year and give their dates. Back each recommended practice with
  evidence that it worked.
- When the user prescribes a practice, look for evidence it works before adopting it.
- Never invent facts, numbers, or reasons. Unknown means unknown: say so.

## Claims

- Verify every claim against the code, including claims from the user. The code is the truth.
- Reason from the known facts of the system before proposing anything new.

## Output

- Present options as a short list showing only what differs: each option, its main trade-off,
  and a recommendation.
- Name the source behind each key fact.
- When the user said to research only, change nothing but the docs they named.

Files in this skill

  • SKILL.md2 KB
  • agents/openai.yaml363 B
  • evals/evals.json4.7 KB
  • evals/trigger-evals.json1.7 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.