Skip to content

datarobot-oss/datarobot-agent-skills / skills/datarobot-agent-assist/agent-assist-deploy-workload

datarobot-agent-assist-deploy-workload

Flagged

Use when an agent assist user wants to deploy their agent, put it on DataRobot, get a link to it, or share it with others, the agent can run as a single container, and the project has no `infra/Pulumi.yaml`. Not for agents built from the agent template, which deploy with Pulumi through datarobot-agent-assist. Users usually just say "deploy" or "deploy to DataRobot" and won't mention containers, workloads or the CLI.

FileKindSize
SKILL.mdskill md10.5 KB
references/app-contract.mdmarkdown3.2 KB
references/deploy-fallback.mdmarkdown7.2 KB
references/deploy-path.mdmarkdown4.4 KB
references/tool-secrets.mdmarkdown1.3 KB

references/app-contract.md

View on GitHub
# Make the app deployable

This skill doesn't build the agent. It deploys the code that's already in `<target_dir>`. If there's no agent code yet, send the user to agent assist option 2 first.

Read the code and check it against the list below. Fix anything that doesn't match yourself, keeping changes to what the deploy needs. Don't rewrite the agent's logic, tools or framework. If the code can't be made to run as one process on one port without rewriting the agent, switch to the Pulumi path in [deploy-path.md](deploy-path.md) and stop.

- One Python project: `pyproject.toml` and a current `uv.lock`. Run `uv lock` after every dependency change. The platform builds from the lockfile.
- Every dependency lives inside `<target_dir>`. A `file://` path or a symlink that leaves it, such as a shared `core` package one folder up, is missing from the image, because the deploy ships only this directory. Check `pyproject.toml` and `uv.lock` for such paths. If the project needs one, switch to the Pulumi path in [deploy-path.md](deploy-path.md) and stop.
- One web server on port 8080 that serves the agent, its API and its UI. Note the entry point (for example `app:app`); SKILL.md section 3 needs it.
- `GET /healthz` returns 200 with no auth and no LLM call. Probes hit the container directly. Add it if it's missing.
- The LLM is called through the LLM Gateway, reading `DATAROBOT_ENDPOINT` and `DATAROBOT_API_TOKEN` from the environment.
- Tool secrets come from env vars. Names go in code, values never do. Follow agent assist's `.env` rules: append `NAME=` to `<target_dir>/.env` and never read the file yourself. Tell the user exactly what to do, for example: "Open the file `.env` in your project folder, find the line `SOME_TOOL_API_KEY=`, paste your key right after the `=`, and save. Don't paste the key here in the chat."
- Serve under the workload prefix. Users reach the app at `/api/v2/endpoints/workloads/<id>/`. The edge strips that prefix before the request reaches the container and doesn't rewrite responses, so every URL the app emits (links, assets, fetch calls, redirects) has to carry it. Build it at startup from the injected `WORKLOAD_ID` and pass it as FastAPI `root_path`:
  ```python
  wid = os.environ.get("WORKLOAD_ID")
  PREFIX = f"/api/v2/endpoints/workloads/{wid}" if wid else ""
  ```
  In the frontend use relative URLs or the prefix. Never a URL starting with `/`. Locally `WORKLOAD_ID` is unset, so the app runs at `/`.
- Who is calling. The edge requires a DataRobot login, then sets `X-Datarobot-User-Id`, `X-Datarobot-Username`, `X-User-Email` and `X-Datarobot-Org-Id` on every request, overwriting whatever the caller sent. Trust those. Don't trust `X-Datarobot-Identity-Token`: a forged value passes through unchanged and there's no published way to verify it. If the spec limits who can use the agent, check `X-User-Email` against an allowlist. Remove any login page.
- The frontend must never send an `Authorization` header to its own backend. The edge treats it as a DataRobot API key and rejects the request before it reaches the container.
- Log to stdout.

Tell the user in one plain sentence what you changed, for example: "I made a few small changes so your agent works on DataRobot."
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall