<!--
CIANCHOSAINT wholesale-copy of cianfhoghlaim/cianfhoghlaim @ main branch.
Original: cianfhoghlaim/cianfhoghlaim (per the openspec/changes/cianchosaint-repo-bootstrap-v2/specs/cianchosaint-bootstrap-v2/spec.md).
Migrated to cianchosaint: 2026-08-23
Licence: BUSL-1.1 (per LICENSE.md)
Wholesale-copied from cianfhoghlaim/cianfhoghlaim@dlthub/SKILL.md.
-->
---
name: dlthub
description: Cianfhoghlaim-side routing skill for the dltHub Platform CLI (`dlthub 1.28+`) — deployment manifest authoring, the `dlthub run` vs `dlthub serve` split, the 5 most common CLI errors and their recovery, and the 5-step `dlthub ai status` health check. Use when the user asks about "dlthub", "deploying to dlthub", "the dlthub CLI", "dlt Hub workspace", "jobs.workspace.dashboard", the `dlthub run`/`dlthub serve` verbs, or the `Matched jobs are interactive` error.
---
# dlthub Platform — Cianfhoghlaim Routing Skill
This skill is the **Cianfhoghlaim-side frontmatter + router** for the dltHub Platform
CLI (`dlthub 1.28+`). The long-form runbook (5-section diagnostic tree,
canonical happy-path workflow, cross-link map) lives at
[`docs/agents/dlthub-run-vs-serve.md`](../../../docs/agents/dlthub-run-vs-serve.md).
The 4 workbench skills below are installed into `.claude/skills/` by
`dlthub ai toolkit install dlthub-platform`.
## When to load
| User says… | Load this skill |
|:--|:--|
| "dlthub run fails with `Matched jobs are interactive`" | Yes (the headline bug) |
| "`dlthub workspace info` shows `No module named 'cianfhoghlaim'`" | Yes |
| "`dlthub ai status` warns about fastmcp" | Yes |
| "How do I deploy a pipeline to dltHub?" | Yes → delegate to `setup-runtime` then `deploy-workspace` |
| "How do I serve the workspace dashboard?" | Yes → `dlthub serve jobs.workspace.dashboard` |
| "Where does the deployment manifest live?" | Yes → `__deployment__.py` |
| "What toolkits should I install?" | Yes → §"Toolkit set" below |
## Decision tree → sub-skill or reference
```text
User wants…
├── "Run a batch pipeline"
│ ├── @run.pipeline("name") in __deployment__.py ← load (deploy-workspace)
│ ├── dlthub run <script.py> ← load (deploy-workspace)
│ └── dlthub local run <job> --dry-run ← load (deploy-workspace)
│
├── "Serve an interactive notebook / dashboard"
│ ├── @run.interactive OR import marimo_module ← load (prepare-deployment)
│ └── dlthub serve <name> ← load (deploy-workspace)
│
├── "Diagnose a CLI error"
│ ├── Interactive-matched error ← runbook §2.1
│ ├── No module named 'cianfhoghlaim' ← runbook §2.2
│ ├── Workspace not yet initialized ← runbook §2.3
│ ├── FastMCP support not installed ← runbook §2.4
│ └── No toolkit with workflow installed ← runbook §2.5
│
└── "Set up / verify the workspace"
├── First-time init ← load (setup-runtime)
├── Production credentials + destination ← load (prepare-deployment)
└── Job failed / flaky ← load (debug-deployment)
```
## Toolkit set (the 8 production toolkits)
Per the runbook §4 + the dlthub-platform skill, the 8 production toolkits
to install for a full-stack setup are:
```bash
dlthub ai init --agent claude # installs `init` (rules + dlt-workspace MCP)
for t in rest-api-pipeline sql-database-pipeline \
filesystem-pipeline dlthub-platform \
data-exploration data-quality transformations; do
dlthub ai toolkit install "$t"
done
uv sync --directory cianfhoghlaim --extra dlthub-platform # installs `fastmcp-slim[server]`
claude # restart so MCP server picks up
```
After install, `dlthub ai status` should list all 8 toolkits and the only
remaining warnings should be the heuristic workspace-marker note (which
disappears once `dlthub init --name cianfhoghlaim` has been run from
inside the workspace root).
## Cross-references
- `docs/agents/dlthub-run-vs-serve.md` — the full diagnostic runbook (5
sections, 5 error recipes, 5-step health check, canonical happy-path)
- `openspec/specs/dlthub-platform-integration/spec.md` — the contract
- `openspec/changes/2026-07-06-wire-dlthub-platform-toolkits-and-deployment/`
— the change record
- `dlthub-ai-workbench/.claude-plugin/marketplace.json` —
the vendored workbench marketplace
- `__deployment__.py` — the deployment manifest
- `dlt/jobs/government_circulars_job.py` — the first
`@run.pipeline("government_circulars_ingest")` job
## Workbench skills (post-install)
After `dlthub ai init + dlthub ai toolkit install dlthub-platform`,
these 4 skills are available to the assistant:
- `setup-runtime` — first-time workspace check (5 steps)
- `prepare-deployment` — production credentials + destination setup
- `deploy-workspace` — `dlthub deploy` / `dlthub local run` / `dlthub
run` / `dlthub serve` / scheduling
- `debug-deployment` — `dlthub job list`, `dlthub job logs`,
`dlthub info`, `dlthub local pipeline list`