@hanv89/azure-arch-skill

CLI installer for the Azure architecture diagram skill. Drops the skill bundle (SKILL.md + worked PlantUML examples) into your AI coding agent's skill folder so that prompts like "draw an Azure context diagram" produce diagrams that use official Microsoft Azure + Fabric architecture icons.

Published with Sigstore provenance via npm Trusted Publishing — verify with npm view @hanv89/azure-arch-skill@latest dist.attestations.

Requirements

• Node.js ≥ 20

Install

Install for one agent:

npx @hanv89/azure-arch-skill install --agent=claude-code
npx @hanv89/azure-arch-skill install --agent=codex
npx @hanv89/azure-arch-skill install --agent=cursor

Install for all supported agents in one shot (transactional — rolls back on first failure):

npx @hanv89/azure-arch-skill install --agent=all

Pin a specific skill bundle version (default = latest from main):

npx @hanv89/azure-arch-skill install --agent=claude-code --version=0.9.0

Supported agents

| Agent | Default install location | Discovery model | |---|---|---| | claude-code | ~/.claude/skills/azure-architecture-diagram/ | Per-user (HOME-relative) | | codex | ~/.codex/skills/azure-architecture-diagram/ (or $CODEX_HOME/skills/...) | Per-user (HOME-relative) | | cursor | <cwd>/.cursor/rules/azure-arch-skill.mdc | Per-project (cwd-relative) — install from the project root so teammates pick up the rule via git |

The cursor adapter prints a one-line note at install time confirming the resolved target; pass --target=<path> to override.

Subcommands

• install --agent=<name> — fetch the latest bundle (or --version=X.Y.Z pinned), verify icon-set reachability + requires_icons semver, copy files into the agent's folder.
• uninstall --agent=<name> — manifest-scoped removal: only the files the install wrote are removed; any user-authored content alongside the skill is preserved.
• update --agent=<name> — short-circuits to a no-op if the on-disk version already matches the upstream bundle; otherwise overwrites.
• list --agent=<name> — list installed skills under the agent's skill root + their versions.
• All four subcommands accept --agent=all to fan out across the supported agents.

Environment overrides

Two environment variables exist for validation / CI use. Production users should not set them.

• AZURE_ARCH_SKILL_BASE_URL — override the bundle source. Restricted to https://raw.githubusercontent.com/hanv89/azure-icons-for-architecture-diagrams/... paths; any other host or scheme is rejected. The CLI emits warn: AZURE_ARCH_SKILL_BASE_URL override active: <value> to stderr whenever the override fires.
• AZURE_ARCH_SKILL_TARGET_ROOT — widen the --target allow-list beyond the agent's default root. Setting this lets install / uninstall / list operate on directories outside the agent's config tree, so it MUST point at a directory you control (e.g. mktemp -d output in a test script). The CLI emits warn: AZURE_ARCH_SKILL_TARGET_ROOT override active: <value> once per command when honored. Never set this in a production shell.

Repository

See the project README for icon source, licensing, and the full skill specification.