octocode-research_skill

This skill performs deep code research across local and remote repos, providing data-driven answers with exact file references and line numbers.

TypeScript

725

GitHub Stars

10

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill bgauryy/octocode-mcp --skill octocode-research

.gitignore207 B
AGENTS.md10.4 KB
ecosystem.config.cjs2.5 KB
eslint.config.mjs684 B
package.json2.3 KB
README.md21.5 KB
SKILL.md22.7 KB
tsconfig.json476 B
tsdown.config.ts970 B
vitest.config.ts449 B

Overview

This skill is an Octocode Research agent for deep code exploration and repository analysis. It performs methodical, phase-driven investigations that return data-backed answers with exact file references and line numbers. Use it to trace flows, find definitions and usages, review PRs, and research external packages or local codebases.

How this skill works

The agent runs a strict sequential process: initialize the local MCP-like server, load context and tool schemas, evaluate whether the request qualifies for a fast-path lookup, plan complex investigations when needed, and then execute iterative research using available tools. It combines LSP-style semantic navigation for local code with GitHub/npm lookups and semantic search tools for external research, always applying self-checks after each action.

When to use it

You need to find where a symbol/function/type is defined or used in a repo
You want to trace execution or call flow across files or services
You need a focused review of a pull request with file/line references
You want to explore an external library, package, or GitHub repo
You need a reproducible research plan for a bug, refactor, or feature

Best practices

Start by ensuring the local MCP server is running before any tool calls
Provide clear, specific targets (symbol names, file paths, or URLs) for fast lookups
Approve the generated research plan for multi-file or multi-repo investigations
Favor fast-path for single-location queries to save time; require planning for cross-domain tracing
Expect iterative tool reads: review hints and confirm before next step

Example use cases

Locate the definition and all usages of formatDate across a TypeScript repo
Trace the authentication flow from API endpoint to token validation and storage
Review a PR URL and summarize critical changes with file/line citations
Compare implementations of a utility across two external npm packages
Prepare a stepwise plan to refactor a module and estimate files affected

FAQ

The agent halts and instructs you to run the server-init script; no research tools are called until the server reports ready.

When is a fast-path used?

Fast-path applies to single-point, unambiguous lookups that expect one file/location and require few tool calls; otherwise the agent builds a plan.

How are results presented?

Findings are data-driven with exact file references and line ranges, plus a summary and actionable next steps.