Task
Every skill in the catalogue whose job is to Automate, whatever stack it runs on.
3 skills
A skill is plain Markdown. The agent loads it on its own when the task matches its description.
/plugin marketplace add https://jesusprodriguez.com/skills/marketplace.json
/plugin install azure-devops@jprodriguez-toolkitThe native route, and the only one that updates itself: add the marketplace once and `/plugin marketplace update` brings in new versions. Skills get their own namespace (`azure-devops:azure-pr-review`).
SKILL=azure-pr-review
curl -fsSL https://jesusprodriguez.com/skills/$SKILL/SKILL.md --create-dirs -o .claude/skills/$SKILL/SKILL.mdOne command at the root of your project. Claude Code discovers it on its own: nothing to register.
$SKILL = "azure-pr-review"
New-Item -ItemType Directory -Force ".claude/skills/$SKILL" | Out-Null
Invoke-WebRequest "https://jesusprodriguez.com/skills/$SKILL/SKILL.md" -OutFile ".claude/skills/$SKILL/SKILL.md"One command at the root of your project. Claude Code discovers it on its own: nothing to register.
curl -fsSL https://jesusprodriguez.com/skills/install.sh -o install.sh
sh install.sh --allDownload the script, read it, run it. Installs one skill, a whole stack or all of them, and pulls in the dependencies between skills. With `--check` it compares the sha256 of what you have installed and lists only what changed.
Automate · Azure DevOps
The access foundation for Azure DevOps: a PAT with the right scope, CLI defaults, WIQL, Analytics OData and honest pagination.
when: When reading from or writing to Azure Repos, Boards or Pipelines from the terminal.
How to ask
Set up Azure DevOps access in this repo and check the PAT has the scopes it needs.
PATs and scopes · az devops / az boards · WIQL queries · REST and pagination
Automate · Azure DevOps
YAML pipelines that do not block the team: one artefact for every environment, versioned templates, secrets kept out of the repo.
when: When creating or changing azure-pipelines.yml, or when a build fails only on the agent.
How to ask
Review azure-pipelines.yml: the build passes locally and fails only on the agent.
Stages and environments · Templates and caching · Secrets and approvals · Debugging the agent
Automate · Python
Automation scripts you can re-run without fear: dry-run, logging, honest exit codes and idempotency.
when: When writing CLIs, scheduled jobs, file processing or API integrations.
How to ask
Write a script that syncs the invoices folder with the SFTP, dry-run by default.
argparse and dry-run · logging · pathlib and encoding · APIs with timeouts