Extended OMO triage: routes to agents, and skills, and tools via keyword scoring for OpenCode.
A fork of cascharly/opencode-triage — the original routes only to skills. This fork extends the dispatch table concept to also route to OMO agents and categories.
npm install -g opencode-triage-omoRestart OpenCode, then type /triage on. That's it.
OpenCode lists every skill's name and description in the prompt on every message. With many skills and OMO agents, that burns hundreds of tokens before you type a word.
Triage hides all skills and agents from the prompt. When you need one, the LLM calls triage() — the plugin finds the best match by keyword scoring and loads only that skill's instructions or routes to the right agent.
No triage: [name+desc of A] [name+desc of B] [name+desc of C] ... [agent list] ... ← every message
With triage: triage({ query }) → keyword match → loads one skill body OR routes to an agent when needed
The more skills and agents you have, the more triage saves. Each new entry adds tokens to the native prompt, but zero tokens when triage is on.
The original opencode-triage only discovers and routes to SKILL.md files. This fork (opencode-triage-omo) extends the routing to also cover:
- Skills — Same as original:
SKILL.mdfiles from.opencode/skills/,.claude/skills/,.agent/skills/,.agents/skills/(project + global) - Agents — OMO agent definitions from
.omo/agents/directories, routing to the right subagent (sisyphus,oracle,explore,librarian, etc.) - Categories — OMO category dispatch for task routing (
visual-engineering,ultrabrain,deep,artistry,quick, etc.) - Tools — Tool selection scoring to determine which MCP tool or built-in tool best fits the query
The scoring engine (keyword + IDF + bigram + phrase + stemming) remains the same powerful deterministic matcher — only the discovery scope and dispatch table are expanded.
The extended triage uses a dispatch table that maps (intent, complexity) pairs to (route_id, agent, tier):
| Intent | Complexity | Route |
|---|---|---|
| UI/UX work | any | visual-engineering category |
| Hard logic | high | ultrabrain category |
| Codebase search | any | explore agent |
| Documentation | any | writing category |
| Security audit | any | security-research skill |
| Git operations | any | git-master skill |
This allows the LLM to be routed to the right agent/category/skill in one triage() call, without having to reason about which subagent type or category to use.
User: "backup my database"
│
▼
LLM: triage({ query: "backup my database" })
│
▼
Plugin: scans filesystem → scores skills + agents → returns best match
backup-restore score=60 (matched: backup, database)
database-sync score=25 (matched: database)
gap=35 ≥ threshold(30) → HIGH CONFIDENCE
No LLM reasoning overhead. No extra API calls. Just fast deterministic matching.
When an OMO agent pattern is matched, the triage result returns structured routing information:
AGENT ROUTED: oracle
Route: oracle (deep reasoning)
Context: agent definition at .omo/agents/oracle.md
The LLM then delegates to the routed agent via the task mechanism, with triage providing the context needed.
If a query word doesn't match any skill, the plugin suggests the closest correction (e.g., scurity → security) and passes it to the LLM as a hint. The LLM can retry silently — no user-facing error.
| Command | What it does |
|---|---|
/triage on |
Enable hooks to hide skills from the LLM prompt (global + local) |
/triage off |
Disable hooks, restore skills to the LLM prompt (global + local) |
/triage status |
See hook state and skill visibility |
/triage dedupe |
Remove project-level duplicates of global skills |
/triage compare |
Token savings estimate for your skills |
See the original opencode-triage README for full command documentation.
npm install -g opencode-triage-omonpm install opencode-triage-omoSame principle as the original — the more skills and agents installed, the more tokens saved per prompt.
- OpenCode 1.14+
- OMO v4.x+
- Bun runtime (bundled with OpenCode)
- Node.js 18+ (for CLI script)
MIT