Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
igorwarzocha avatar

Agent Navigation Sop

  • 83 installs
  • 125 repo stars
  • Updated February 4, 2026
  • igorwarzocha/opencode-workflows

Generate a dense AGENTS.md for AI agent navigation covering build/test commands, coding conventions, and task routing during /init.

About

An SOP that complements OpenCode's /init to produce developer-facing AGENTS.md navigation docs. A developer uses it to map build/test/lint commands, forbid blocking processes, and route common tasks to directories.

  • Targets ~50 lines: build & test commands, constraints & patterns, task routing
  • Explicitly forbids dev servers/watch modes and flags read-only paths

Agent Navigation Sop by the numbers

  • 83 all-time installs (skills.sh)
  • +3 installs in the week ending Aug 2, 2026 (Skillselion tracking)
  • Ranked #681 of 1,879 Documentation skills by installs in the Skillselion catalog
  • Data as of Aug 2, 2026 (Skillselion catalog sync)
npx skills add https://github.com/igorwarzocha/opencode-workflows --skill agent-navigation-sop

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs83
repo stars125
Last updatedFebruary 4, 2026
Repositoryigorwarzocha/opencode-workflows

What it does

Generate a dense AGENTS.md for AI agent navigation covering build/test commands, coding conventions, and task routing during /init.

Files

SKILL.mdMarkdownGitHub ↗

Agent Navigation SOP

This SOP COMPLEMENTS the /init command by providing deep-dive mapping for developer-facing AGENTS.md files.

<workflow>

Phase 1: Deep Analysis

The agent SHOULD skip basic directory scans (opencode already sees the tree) and focus on:

1. Tooling & Scripts: Identify one-shot commands for Build, Test, Lint, and Typecheck. 2. Process Constraints: Identify dev servers and watch-modes to EXCLUDE. 3. Conventions: Identify project-specific naming, architectural patterns, and unique error handling.

Phase 2: AGENTS.md Composition

Target a high-density output of ~50 lines.

1. Build & Test (Instructions)

  • List copy-pasteable, one-shot commands.
  • MUST NOT include dev, watch, or interactive commands.

2. Constraints & Patterns (Rules)

  • CRITICAL: The agent MUST explicitly forbid running blocking processes (dev servers, watch modes).
  • Identify read-only or restricted paths that agents SHOULD NOT modify.
  • Document repository-specific coding patterns (e.g., "Use Bun APIs", "Prefer functional over class").

3. Task Routing (Routing)

  • Map common tasks (feature addition, bug fix) to specific directories and related tests.
  • Reference nested AGENTS.md files for package-level details instead of duplicating them.
  • Identify repetitive tasks and SHOULD suggest creating new skills using skill-creator to extend agent capabilities.

</workflow>

<instructions>

  • The agent MUST NOT repeat environment metadata (date, OS, absolute paths).
  • Output length of ~50 lines is STRONGLY RECOMMENDED.
  • The agent MUST use RFC 2119 keywords (MUST, MUST NOT, SHOULD, STRONGLY RECOMMENDED).
  • The agent MUST prioritize "Access Constraints" (read-only files/folders).
  • The agent MUST NOT document long-running/blocking processes.
  • The agent MUST document one-shot verification commands (build, lint, test, typecheck, codegen).
  • The agent MUST NOT include prose that does not change agent behavior.

</instructions>

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.