speckit-infrahub-speckit-route-implement

name	speckit-infrahub-speckit-route-implement
description	Hook command — runs before the active integration's speckit implement command in Infrahub projects. Invokes the matching infrahub-managing-* skill for each artifact type that tasks.md touches, so implementation code follows authoritative Infrahub patterns and conventions.
compatibility	Requires spec-kit project structure with .specify/ directory
metadata	{"author":"github-spec-kit","source":"infrahub-speckit:commands/speckit.infrahub-speckit.route-implement.md"}

Infrahub Routing — pre-implement hook

This command is fired as a before_implement hook by the infrahub-speckit extension. It runs every time the active integration invokes the core implement workflow, for example /speckit-implement in Codex skills or /speckit.implement in Claude Code slash commands. If .infrahub.yml is not present, this command is a no-op.

MANDATORY — do NOT skip, defer, or rationalize around this.

Before any task that touches Infrahub artifacts is implemented, the corresponding Infrahub skill must be loaded. The skill loads authoritative reference material, validation rules, and implementation patterns that the code MUST follow.

Hook return semantics

Throughout this file, "return" means: stop emitting hook-command output and let control flow back to the calling core skill. Do NOT terminate the session, abort the parent slash command, or skip the core skill body. The spec-kit hook contract requires this hook to either:

• emit a no-routing line and return (no-op case), OR
• emit the Step 5 directive block and return (routing case), OR
• emit a user-facing error message and halt the entire command sequence (preflight failure case — the calling skill should NOT proceed in this case).

The third case is the only one where the calling skill should not resume.

Step 1 — Detect Infrahub project

If the repository has no .infrahub.yml, emit:

[infrahub-speckit] No .infrahub.yml detected. No routing applied.

Then return. The core implement skill runs unchanged.

Step 2 — Verify required Infrahub skills are installed

Confirm these skills appear in your available-skills inventory:

• infrahub-managing-schemas
• infrahub-managing-transforms
• infrahub-managing-checks
• infrahub-managing-generators
• infrahub-managing-menus
• infrahub-managing-objects

If ANY are missing, halt and tell the user:

The infrahub-speckit extension requires the OpsMill Infrahub skills for the
active agent.

Install or enable the `infrahub-managing-*` skills for this environment.

For Codex sessions, confirm the skills are listed in the available skills
inventory. For Claude Code sessions, install the skills package or plugin:
  npx skills add opsmill/infrahub-skills
  /plugin marketplace add opsmill/claude-marketplace
  /plugin install infrahub@opsmill

Docs: https://docs.infrahub.app/skills/installation-setup

After installing, restart this session and re-run the implement command.

Do NOT proceed further until the skills are installed. Do NOT continue to later steps or emit any routing directive — exit the hook with the user-facing error message only.

Step 3 — Identify artifact types touched by tasks.md

1. Read tasks.md and plan.md from the current feature directory.

2. Identify every artifact type any task in any phase touches:

   Task Involves	Skill to Invoke	When to Invoke
   Schema YAML files (schemas/)	infrahub-managing-schemas	Before the first schema task
   Python transforms (transforms/)	infrahub-managing-transforms	Before writing transform code
   Python checks	infrahub-managing-checks	Before writing check code
   Generator Python files (generators/)	infrahub-managing-generators	Before writing generator code
   Menu YAML files (menus/)	infrahub-managing-menus	Before writing menu definitions
   Object data files (objects/)	infrahub-managing-objects	Before writing seed data

3. Build the list of skills to load — one per artifact type touched, deduplicated.

Step 4 — Invoke each matched skill

Invoke each skill on the deduplicated list using the Skill tool, BEFORE returning. Each skill loads once per implementation session — no need to re-invoke per task within the same session.

Anti-rationalization check — if you think "I already loaded the skill during specify" or "I already loaded it during plan" or "the plan has all the details I need", invoke the skill anyway. Skill content is NOT persisted across commands; each command session starts fresh.

Step 5 — Emit the routing record and return

Emit a single status line listing the skills loaded:

[infrahub-speckit routing — applies to this implement run only]

Skills loaded: <comma-separated list>

Then return. The core speckit-implement skill runs next.