research

Investigate external documentation and vendor APIs, then capture cited findings as Markdown files.

2|Updated Jul 24, 2026
One-click install
npx skills add https://github.com/konstantinos-malavazos/claude-code-playbook --skill research-konstantinos-malavazos
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: research
Source: https://github.com/konstantinos-malavazos/claude-code-playbook/tree/main/templates/skills/research
Command: npx skills add https://github.com/konstantinos-malavazos/claude-code-playbook --skill research-konstantinos-malavazos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When a decision is blocked on a fact nobody on the team knows — a third-party API behavior, a spec detail, an RFC requirement — this Skill turns that unknown into a cited, verifiable finding without derailing the main work session. ## Core Features & Use Cases - Background dispatch: Hands the research question to a background agent so the main session keeps working while the investigation runs. - Primary-source discipline: Requires every claim to be traced to official docs, specs, or vendor references, with citations and versions recorded. - Negative-result reporting: Captures "the docs do not say" and "X is not supported" as first-class findings so they are never re-researched. - Memory integration: Banks durable conclusions (rate limits, auth quirks, ruled-out approaches) as tagged memories, and queries memory first to avoid duplicate investigations. - Use Case: You need to know whether a payment provider's webhook retries on HTTP 500 and with what backoff. Dispatch the question, keep coding, and receive a cited Markdown findings file plus a clear owner for the decision it unblocks. ## Quick Start Ask the research skill to find out from the vendor's official documentation whether provider X retries failed webhooks and to write the cited findings to a Markdown file.

Frequently Asked Questions about research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I research a third-party API question without losing my place?

Dispatch the question to a background agent and keep working in the main session. Give the agent a precise question, such as whether a webhook retries on a 500 and with what backoff, rather than a broad topic.

When should I use external research instead of code search?

Use research only for knowledge outside the workspace, like vendor APIs, specs, and RFCs. Questions about your own code's symbols or callers belong to semantic code navigation tools, and previously concluded findings should be queried from memory first.

What makes a research finding trustworthy enough to decide on?

Every claim must be traced to a primary source such as official docs or the vendor's own API reference, with a link and the version read. A fact without a citation is treated as a guess, not a finding.

Does the research flow handle cases where the answer does not exist?

Yes. Negative results like "the docs do not say" or "X is not supported" are recorded as findings. The agent must also state why the source cannot answer and what would resolve it, such as a spike or support ticket.

What happens to research findings after the investigation finishes?

In-flight findings sit with the ticket's handoffs, while durable conclusions like rate limits or ruled-out approaches are banked as memories tagged by topic or component, never by ticket id. Findings are handed to whoever owns the decision.