Agent Help
I'm using the agent-help skill to explain the agent pipeline system.
Quick Reference
Available Commands
| Command |
Agent |
What It Does |
/code |
Coding Agent |
Full-stack builds, quality audits, prototypes, debugging |
/code full-stack |
Coding Agent |
brainstorm → plan → parallel build → review → finish |
/code quality |
Coding Agent |
code-review → bugs → security → tests → perf → verify |
/code prototype |
Coding Agent |
api → db → FastAPI → React → Docker |
/code debug |
Coding Agent |
debug → root-cause → fix → test → verify |
/docs |
Docs Agent |
API docs, project scaffolding, content, releases |
/docs api |
Docs Agent |
documenter → code-docs → examples → OpenAPI |
/docs project |
Docs Agent |
CLAUDE.md → ARCHITECTURE → DEVLOG → DOMAIN |
/docs content |
Docs Agent |
research → parallel(Medium, LinkedIn, newsletter) → polish |
/docs release |
Docs Agent |
PRs → commits → changelog → release-notes |
/startup |
Startup Agent |
MVPs, go-to-market, pitch decks, analytics |
/startup mvp |
Startup Agent |
brainstorm → Hormozi → plan → parallel build → deploy |
/startup gtm |
Startup Agent |
research → pitch → parallel(5 channels) → email |
/startup pitch |
Startup Agent |
Hormozi → coach → PPTX → outreach email |
/startup analytics |
Startup Agent |
metrics → pipeline → D3.js → PPTX report |
How It Works
- Say
/code, /docs, or /startup (optionally with a pipeline name)
- The super-agent asks what you want to build and which mode to use
- It loads the pipeline JSON and invokes the pipeline-orchestrator
- The orchestrator runs each stage, passing outputs between them
- Gates pause for approval, run code review, or require verification
- On completion, you get a summary of all outputs and next steps
Modes
- Interactive (default): Pauses at every gate for your approval
- Autonomous: Skips approval gates, still enforces quality and verification
Creating Custom Pipelines
Read the full guide: agents/GUIDE.md
- Create a JSON file in
agents/<agent>/references/pipelines/<name>.json
- Define stages with
skill, skill-chain, or parallel types
- Add gates (
approval, quality, verification) between stages
- Declare
outputs and inputs for data passing
- Update the agent's SKILL.md routing table
Troubleshooting
- Pipeline won't start: Check SKILL.md exists, JSON is valid, restart session
- Stage fails: Check skill path exists, verify input files are on disk
- Parallel conflict: Ensure branches write to different output paths
- Gate blocks: Fix Critical issues from quality gate, provide evidence for verification
For the complete guide with architecture diagrams, JSON schema, examples, and custom pipeline tutorial, read agents/GUIDE.md.
1---2name: agent-help3description: Use when the user asks about agent pipelines, how to use /code /docs /startup commands, how to create custom pipelines, or needs help troubleshooting agent workflows - explains the agent orchestrator system and guides pipeline creation4---56# Agent Help78I'm using the `agent-help` skill to explain the agent pipeline system.910## Quick Reference1112### Available Commands1314| Command | Agent | What It Does |15|---------|-------|-------------|16| `/code` | Coding Agent | Full-stack builds, quality audits, prototypes, debugging |17| `/code full-stack` | Coding Agent | brainstorm → plan → parallel build → review → finish |18| `/code quality` | Coding Agent | code-review → bugs → security → tests → perf → verify |19| `/code prototype` | Coding Agent | api → db → FastAPI → React → Docker |20| `/code debug` | Coding Agent | debug → root-cause → fix → test → verify |21| `/docs` | Docs Agent | API docs, project scaffolding, content, releases |22| `/docs api` | Docs Agent | documenter → code-docs → examples → OpenAPI |23| `/docs project` | Docs Agent | CLAUDE.md → ARCHITECTURE → DEVLOG → DOMAIN |24| `/docs content` | Docs Agent | research → parallel(Medium, LinkedIn, newsletter) → polish |25| `/docs release` | Docs Agent | PRs → commits → changelog → release-notes |26| `/startup` | Startup Agent | MVPs, go-to-market, pitch decks, analytics |27| `/startup mvp` | Startup Agent | brainstorm → Hormozi → plan → parallel build → deploy |28| `/startup gtm` | Startup Agent | research → pitch → parallel(5 channels) → email |29| `/startup pitch` | Startup Agent | Hormozi → coach → PPTX → outreach email |30| `/startup analytics` | Startup Agent | metrics → pipeline → D3.js → PPTX report |3132### How It Works33341. Say `/code`, `/docs`, or `/startup` (optionally with a pipeline name)352. The super-agent asks what you want to build and which mode to use363. It loads the pipeline JSON and invokes the pipeline-orchestrator374. The orchestrator runs each stage, passing outputs between them385. Gates pause for approval, run code review, or require verification396. On completion, you get a summary of all outputs and next steps4041### Modes4243- **Interactive** (default): Pauses at every gate for your approval44- **Autonomous**: Skips approval gates, still enforces quality and verification4546### Creating Custom Pipelines4748Read the full guide: `agents/GUIDE.md`49501. Create a JSON file in `agents/<agent>/references/pipelines/<name>.json`512. Define stages with `skill`, `skill-chain`, or `parallel` types523. Add gates (`approval`, `quality`, `verification`) between stages534. Declare `outputs` and `inputs` for data passing545. Update the agent's SKILL.md routing table5556### Troubleshooting5758- **Pipeline won't start:** Check SKILL.md exists, JSON is valid, restart session59- **Stage fails:** Check skill path exists, verify input files are on disk60- **Parallel conflict:** Ensure branches write to different output paths61- **Gate blocks:** Fix Critical issues from quality gate, provide evidence for verification6263For the complete guide with architecture diagrams, JSON schema, examples, and custom pipeline tutorial, read `agents/GUIDE.md`.
ViggyV