omo_skill

This skill routes and coordinates multi-agent work flows for code analysis, bug investigation, fix planning, and implementation to minimize changes.
  • Go

2.1k

GitHub Stars

2

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 cexll/myclaude --skill omo

  • README.md4.0 KB
  • SKILL.md8.8 KB

Overview

This skill orchestrates multi-agent workflows for code analysis, bug investigation, fix planning, and implementation. It selects the minimal set of agents and ordering based on task type and risk, then invokes them with full context. Use the /omo trigger to let the orchestrator (Sisyphus) delegate discovery, analysis, and implementation steps without writing code itself.

How this skill works

Sisyphus inspects the user request and routing signals to decide which agents to call (explore, librarian, oracle, develop, frontend-ui-ux-engineer, document-writer). It issues each agent call via the prescribed codeagent-wrapper format, always including the original request and a Context Pack containing prior agent outputs. The orchestrator never modifies code directly and prefers the fewest agents needed to meet acceptance criteria.

When to use it

  • You need a staged approach: locate problem, analyze, then implement.
  • Location of bug or change is unknown and you need code discovery.
  • External library/API usage or compatibility must be researched.
  • Change is multi-file, public API–affecting, or otherwise high risk.
  • You want a minimal, auditable agent plan rather than a one-shot code edit.

Best practices

  • Always include the original user request in every agent invocation.
  • Pass complete Context Packs: explore, librarian, and oracle outputs when available.
  • Prefer the smallest agent set that satisfies acceptance criteria; skip steps when signals do not apply.
  • Delegate all code changes to an implementation agent; never write code yourself in the orchestrator.
  • Use oracle for tradeoffs and high-risk multi-file changes, not as a default pipeline.

Example use cases

  • Explain how a function works: run explore only to produce a call-chain summary and snippets.
  • Unknown-location bug fix: explore to locate root cause, then develop to implement a narrow fix.
  • Add a new feature using an external library: run explore (internal) and librarian (external) in parallel, then oracle for planning, then develop to implement.
  • Cross-cutting refactor: explore to map touches, oracle for risk and plan, develop to apply changes and tests.

FAQ

Invoke oracle when changes are risky: multi-file impacts, public API changes, data format or config changes, concurrency/security/perf concerns, or unclear tradeoffs.

Can I skip explore if I already gave a file and line?

Yes. Skip explore when the user provided an exact file path and line or the context already includes precise locations.

Who runs tests and applies patches?

An implementation agent (develop or a specialized frontend/document writer) performs edits and runs the narrowest relevant tests. The orchestrator only delegates and supplies context.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational