discover

Maps repository structure, API surfaces, and governance rules through read-only exploration.

Updated May 31, 2026
One-click install
npx skills add https://github.com/AlexanderNarbaev/agi --skill discover-alexandernarbaev
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: discover
Source: https://github.com/AlexanderNarbaev/agi/tree/main/.opencode/skills/discover
Command: npx skills add https://github.com/AlexanderNarbaev/agi --skill discover-alexandernarbaev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Before writing or modifying code, developers need a clear picture of the repository: its structure, public API surfaces, complexity hotspots, and project governance rules. This Skill automates that discovery phase so implementation work starts from verified context instead of guesswork. ## Core Features & Use Cases - Explorer Delegation: Delegates read-only repository discovery to the swarm's explorer agent, with an optional second pass focused on hidden requirements, unstated assumptions, and scope risks. - API Surface Analysis: Runs symbols and batch_symbols tools on key files to map public interfaces, plus complexity_hotspots to flag modules needing security review or full gates. - Governance Extraction: Locates project governance files (project-instructions.md, CONTRIBUTING.md, INSTRUCTIONS.md) in priority order, extracts MUST and SHOULD rules, and writes them to .swarm/context.md under a Project Governance section. - Use Case: Before implementing a new feature in an unfamiliar monorepo, run the discover protocol to collect API surfaces, risk areas, and mandatory contribution rules into a shared context file for the whole agent swarm. ## Quick Start Ask the architect agent to run MODE: DISCOVER on this repository so it maps the codebase, API surfaces, and governance rules before any implementation begins.

Frequently Asked Questions about discover

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

FAQPage Schema
How do I explore an unfamiliar codebase before making changes?

Delegate discovery to an explorer agent that reads the repository without modifying it, then run symbols analysis on key files to map public API surfaces. A second explorer pass focused on risks and hidden requirements catches unstated assumptions before implementation starts.

What is the difference between symbols and batch_symbols for code analysis?

The symbols tool inspects the public API surface of individual files, while batch_symbols processes multiple files in one call. For multi-file module surveys, batch_symbols is preferred over sequential single-file symbols calls to reduce overhead.

How do I extract coding rules from CONTRIBUTING.md files?

Locate the governance file using glob patterns in priority order: project-instructions.md, docs/project-instructions.md, CONTRIBUTING.md, then INSTRUCTIONS.md. Read the first match and extract all MUST (mandatory) and SHOULD (recommended) rules into a summary section.

Does the discover protocol modify any repository files?

The discovery itself is read-only; the explorer agent only reads repository content. The single write operation is appending extracted governance rules to the .swarm/context.md file under a Project Governance section.

When should complexity_hotspots analysis be run during discovery?

Run complexity_hotspots during discovery only if it was not already executed in Phase 0; check context.md for existing analysis first. Modules flagged with security_review or full_gates recommendations should be noted in context.md for later phases.