name	code-research
description	Research open-source repositories to understand how something is built or works.
metadata	{"author":"morph","version":"0.1.0","argument-hint":"<question about an external repo>"}

Code Research

Use github_codebase_search to explore external open-source repositories and understand their implementation details.

When To Use

The user asks how an open-source library or framework implements something.
You need to understand an external dependency's internals before integrating with it.
The user wants to compare how different repos solve a problem.
You need to find usage examples or patterns in a public GitHub repo.

Steps

Identify the GitHub repository (e.g. vercel/next.js, facebook/react).
Break the user's question into multiple focused search angles (e.g. entry points, data flow, config, tests).
Fire multiple github_codebase_search calls in parallel — one per angle. Do not run them sequentially.
Read and cross-reference the returned files/sections from all queries.
Summarize findings with concrete file paths and code references.

Parallel Queries

Always decompose the research into 2-4 concurrent searches. For example, to understand how a repo handles authentication:

Query 1: "Find the authentication middleware and session handling"
Query 2: "Find the login/signup API routes and handlers"
Query 3: "Find the auth configuration and token validation"

Run all queries at the same time to minimize latency.

Query Template

"Find how implements . Include entry points, key functions, and data flow."

Plus depuis ce dépôt

même dépôt

browser-use

marktantongco/opencode-accomplishments

Headless and headful browser automation with natural language control. Saves tokens by delegating all Playwright logic to a dedicated browser automation layer.

2026-05-270

agent-roles

marktantongco/opencode-accomplishments

Unified multi-agent role system. Covers Builder, Coder, Council, Orchestrator, Plan, Researcher, Reviewer, and Scribe personas. Use when coordinating multi-agent workflows with MECE principle and role delegation.

2026-05-060

anthropic-frontend-design

marktantongco/opencode-accomplishments

Bold aesthetic principles for AI-native interfaces, conversational UI, and trust-forward design. Use when building Claude-like interfaces, trust signals, conversational layouts, or intellectual generosity patterns.

2026-05-060

audit-analyzer

marktantongco/opencode-accomplishments

Automates detection and prioritization of audit-related issues across the workspace. Scans for performance, accessibility, and monitoring signals. Use when analyzing audit errors or requesting actionable improvement patches.

2026-05-060

explore

marktantongco/opencode-accomplishments

Use this skill when the user asks you something that requires searching a code base

2026-05-060

feature-research

marktantongco/opencode-accomplishments

Research existing architecture before implementing a complex feature.

2026-05-060

name	code-research
description	Research open-source repositories to understand how something is built or works.
metadata	{"author":"morph","version":"0.1.0","argument-hint":"<question about an external repo>"}

Code Research

Use github_codebase_search to explore external open-source repositories and understand their implementation details.

When To Use

The user asks how an open-source library or framework implements something.
You need to understand an external dependency's internals before integrating with it.
The user wants to compare how different repos solve a problem.
You need to find usage examples or patterns in a public GitHub repo.

Steps

Identify the GitHub repository (e.g. vercel/next.js, facebook/react).
Break the user's question into multiple focused search angles (e.g. entry points, data flow, config, tests).
Fire multiple github_codebase_search calls in parallel — one per angle. Do not run them sequentially.
Read and cross-reference the returned files/sections from all queries.
Summarize findings with concrete file paths and code references.

Parallel Queries

Always decompose the research into 2-4 concurrent searches. For example, to understand how a repo handles authentication:

Query 1: "Find the authentication middleware and session handling"
Query 2: "Find the login/signup API routes and handlers"
Query 3: "Find the auth configuration and token validation"

Run all queries at the same time to minimize latency.

Query Template

"Find how implements . Include entry points, key functions, and data flow."