# Heyeddi Orchestrator

> Discover HeyEddi skills, load the catalog, and suggest which @skills to invoke for the current task. Use at session start, ambiguous requests, multi-skill pipelines, or when the user asks what skills are available. Reads skill-routing.json from @heyeddi-intake when present.

- Skill: `heyeddi-com/heyeddi-orchestrator-3` (Agent Skill, multi-file: 14 files)
- Install (CLI): `npx skillmds@latest add heyeddi-com/heyeddi-orchestrator-3`
- Raw SKILL.md: https://api.skillmd.com/api/skills/heyeddi-com/heyeddi-orchestrator-3/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Integrations & APIs
- Author: HeyEddi-com (https://skillmd.com/u/heyeddi-com)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/heyeddi-com/heyeddi-orchestrator-3

---


# Skill Orchestrator

Routes work to the right HeyEddi skill — without loading every SKILL.md at once.

## When to use

- Session start on a HeyEddi project (`.heyeddi/` present or greenfield app request)
- User asks "what skills do we have?" or "which skill should handle this?"
- Before a multi-step pipeline (intake → scaffold → design → handoff → QA)
- After `@heyeddi-intake` — confirm `skill-routing.json` before downstream work

## Limits (Cursor)

Cursor **cannot** inject all skill bodies automatically. This skill:

1. **`write_skills_index`** — scan once → `.heyeddi/skills-index.json` + `skills-index.md`
2. **`load_catalog`** — read the cached index (rescan with `--refresh` after skill installs)
3. **`suggest_skills`** — ranked matches from index entries (+ `skill-routing.json` when present)
4. **You** — read **one** chosen skill's `SKILL.md`, then invoke it

## Pipeline

```
write_skills_index        → .heyeddi/skills-index.{json,md}  (session start / after installs)
read skills-index.md      → quick catalog without opening every SKILL.md
suggest_skills            → ranked @skills for this task
read one SKILL.md         → follow that skill's pipeline
```

If `.heyeddi/docs/intake/skill-routing.json` exists (from `@heyeddi-intake`), **follow route order** — suggestions surface it at score 100.

## Instructions

1. **Session start:** `write_skills_index --project-root <root>` (or `load_catalog --refresh` — same effect)
2. **Quick lookup:** read `.heyeddi/skills-index.md` — no need to open every SKILL.md
3. `suggest_skills --project-root <root> --user-prompt "<request>"`  
   Or `--prompt-file USER_PROMPT.md`
4. Read **one** chosen skill's `SKILL.md` (path in index JSON), then invoke `@skill-name`
5. Re-run `write_skills_index` after `npx skills add` or hub skill updates
6. Greenfield with no `.heyeddi/product.md` → top suggestion often `@heyeddi-intake`

## Tools

| Script | Purpose |
|--------|---------|
| `write_skills_index.py` | Scan → `.heyeddi/skills-index.*` |
| `load_catalog.py` | Read index (or scan+write if missing) |
| `suggest_skills.py` | Rank skills from cached index |

## Related

- `@heyeddi-intake` — writes `skill-routing.json`
- `reference/skill-triggers.md` — how per-skill `reference/triggers.md` works (fully agnostic scoring)