service-cli-registry

Generated source view for the actual executable engineering/service-cli-registry skill. The durable routing article is Cloud, Data, and Service Skills. Source: skills/engineering/service-cli-registry/SKILL.md

Runtime Source

Bundled Resources

No bundled resource files.

Description

Route SaaS and infrastructure tasks to the right first-party CLI, MCP, or installed skill. Use when a task mentions GitLab, AWS, Google Cloud, Docker, Kubernetes, Terraform, Sentry, Datadog, Snowflake, Databricks, Auth0, Twilio, Contentful, Sanity, Algolia, or any SaaS/infra product not already covered by a more specific skill. Enforces auth/context checks, read-only-first exploration, dry-run/plan/preview before mutations, and capability-harvest lookup for unknown services.

Skill Source

---
name: service-cli-registry
description: Route SaaS and infrastructure tasks to the right first-party CLI, MCP, or installed skill. Use when a task mentions GitLab, AWS, Google Cloud, Docker, Kubernetes, Terraform, Sentry, Datadog, Snowflake, Databricks, Auth0, Twilio, Contentful, Sanity, Algolia, or any SaaS/infra product not already covered by a more specific skill. Enforces auth/context checks, read-only-first exploration, dry-run/plan/preview before mutations, and capability-harvest lookup for unknown services.
origin: personal
source_slug: service-cli-registry
---

# Service CLI Registry

Use this skill when the user asks for work involving a SaaS or infrastructure product and no more specific installed skill already owns the task.

This is the routing layer for `config/capability-harvest-seeds.json` and `wiki/tools/service-cli-coverage.md`. It turns the harvested registry into agent behavior without installing every MCP or bespoke skill.

## First Action

1. Check for a more specific installed skill:
   - `rg --files skills/personal | rg '<service>|<domain>'`
   - `qmd search "<service> skill mcp cli"`
2. If a service-specific skill exists, use it first.
3. If not, use `wiki/tools/service-cli-coverage.md` and `config/capability-harvest-seeds.json` to identify the preferred interface.
4. If the service is missing, search the latest harvest:
   - `npm run promote:capabilities -- --limit 25`
   - `rg -i "<service>" raw/registries/agent-capabilities config/capability-harvest-seeds.json`

## Interface Order

Prefer the lowest-friction interface that can do the job:

| Rank | Interface | Use when |
| --- | --- | --- |
| 1 | Specific installed skill or plugin MCP | The task matches an installed capability and the tool is available. |
| 2 | First-party CLI | The CLI has a stable documented surface and the task is shell-friendly. |
| 3 | Existing PrintingPress CLI | Repeated API/data work benefits from local SQLite sync + search/sql. |
| 4 | Official MCP / Smithery candidate | The agent runtime needs MCP and the server has credible source/auth metadata. |
| 5 | Raw HTTP/API | Last resort after checking docs and credentials. |

Do not install MCPs from the raw harvest just because they exist. Treat MCP candidates as leads that need auth, permission, and owner review.

## Safety Gates

Before any SaaS/infra command:

1. **Identify account/context.** Print the current account, org, subscription, project, region, cluster, namespace, or tenant before using it.
2. **Read first.** Prefer list/show/status/log commands before create/update/delete.
3. **Use previews.** Run `--dry-run`, `plan`, `preview`, `diff`, `validate`, `template`, or equivalent before mutation.
4. **Ask before destructive or expensive changes.** This includes deletes, prod deploys, infra applies, paid resource creation, broad permission changes, email/message sends, and anything involving user/customer data.
5. **Prefer scoped credentials.** Do not ask for global admin tokens when a read-only or project-scoped token works.

## High-Priority Missing Areas Covered

Use this skill to cover recurring gaps that do not need their own local skill yet:

- Developer platforms: GitLab via `glab`; GitHub defaults to `gh`.
- Clouds: AWS via `aws`, Google Cloud via `gcloud`/`bq`; Azure has dedicated personal skills.
- Infra: Docker, Kubernetes, Helm, Terraform, Pulumi, Ansible.
- Observability: Sentry CLI, Datadog CI, New Relic CLI, Grafana CLI.
- Data and warehouses: PostgreSQL, MySQL, MongoDB, ClickHouse, Snowflake, Databricks, BigQuery.
- Product SaaS: Auth0, Okta, Twilio, SendGrid, Contentful, Sanity, Strapi, Algolia.

## Verification

After using a service CLI, capture:

- command(s) run
- account/context observed
- whether the command was read-only, dry-run, or mutating
- artifact/result path if one was generated

For recurring successful patterns, update `wiki/tools/service-cli-coverage.md` or create a service-specific skill.

## Related Skills

- `find-skills` - discover upstream skills before hand-rolling.
- `skill-auditor` - audit external skills before install.
- `wrangler`, `supabase`, `clerk-setup`, `firebase-basics`, `azure-*`, `gws-*` - service-specific owners where already installed.

---

Timeline

• 2026-07-15 | Generated a browseable source page from the actual executable skill file. Source: skills/engineering/service-cli-registry/SKILL.md