mhd can run deploy/tasks.py

.claude/skills/update-deploy-py/SKILL.md

@@ -0,0 +1,63 @@
+---
+name: update-deploy-py
+description: This skill should be used when the user asks to "update deploy-py", "port deploy-js to deploy-py", "sync deploy-py", "reconcile the python port", or wants to bring the deploy-py Python port back in line with this repo (deploy-js, the source of truth) and open a PR for any drift.
+---
+
+# Update deploy-py from deploy-js
+
+This repo (`deploy-js`, TypeScript) is the source of truth. `deploy-py` is a
+Python port of it. When deploy-js changes, port the change into deploy-py and
+open a PR there. This runs both in CI (push to `master`) and locally for a
+developer driving Claude Code by hand.
+
+## Get latest deploy-py
+
+Always clone the latest default branch (`main`) of deploy-py into a fresh temp
+dir. Do not reuse a local sibling checkout, which may be stale. deploy-py is the
+target you edit; this repo (deploy-js) is read-only reference.
+
+```bash
+DEPLOY_PY=$(mktemp -d)
+gh repo clone mindhivenz/deploy-py "$DEPLOY_PY" -- --depth 1 --branch main
+```
+
+`gh` uses `GH_TOKEN` in CI and the developer's auth locally.
+
+## Task
+
+1. Identify what changed in deploy-js. In CI, look at the latest push to
+   `master` (`git log -p -1`, or diff against the previous commit). Locally,
+   review recent commits the developer points you at. Focus on `src/`.
+2. Map each changed deploy-js source file/function to its deploy-py counterpart
+   under `$DEPLOY_PY/src/mindhivenz/deploy/`. Naming converts TS camelCase to
+   Python snake_case (e.g. `userRoleName.ts` -> `user_role_name.py`).
+3. If deploy-py already reflects the change, STOP. Do not open a PR, do not
+   commit, do not create a branch. Print a short note that deploy-py is in sync
+   and exit.
+4. Before doing any work, check deploy-py for an already-open PR from a previous
+   run (branch prefix `deploy-js-sync/` or PR title starting with
+   "Sync from deploy-js"). If one exists and already covers the change, STOP and
+   do nothing.
+5. If there IS new drift to apply: in `$DEPLOY_PY`, create a branch named
+   `deploy-js-sync/<short-date-or-topic>`, edit only files under `$DEPLOY_PY`
+   (never anything in this deploy-js repo) to match the deploy-js change, then
+   run the tests with `uv run pytest`. Use `uv` / `uv run` for all Python
+   commands.
+6. Commit and open a PR against `main` on deploy-py, requesting `timvan` as a
+   reviewer (`gh pr create --reviewer timvan`). If running locally as `timvan`
+   (you cannot request review from yourself), omit the flag. In the PR body, list
+   each ported change as: the deploy-js source (file + function), the deploy-py
+   target file, what changed, and the reason. Link the specific deploy-js commit
+   and lines on GitHub. Do not merge.
+
+## Constraints
+
+- deploy-js (this repo) is reference only. Never modify, commit to, or push it.
+- Keep edits minimal and faithful to deploy-js. Port behavior, do not refactor
+  unrelated deploy-py code or invent functionality not present in deploy-js.
+- Match deploy-py's existing idioms (typer, snake_case, project layout). Do not
+  transliterate TypeScript style into Python.
+- Ignore changes confined to deploy-js build artifacts (`dist/`), tooling, or
+  packaging that have no behavioral equivalent in deploy-py.
+- If tests fail and you cannot make them pass with a faithful port, open the PR
+  anyway as a draft and describe the failure in the body.

.github/workflows/sync-deploy-py.yaml

@@ -0,0 +1,53 @@
+name: Sync deploy-py from deploy-js
+
+# Runs on every push to master (this repo, deploy-js, is the source of truth) and
+# asks Claude Code to port any source changes that have not yet been reflected in
+# the deploy-py Python port. If it finds divergence, it opens a PR on deploy-py.
+on:
+  push:
+    branches:
+      - master
+  workflow_dispatch: {}
+
+permissions:
+  contents: read
+
+jobs:
+  sync-deploy-py:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Generate GitHub App token
+        id: app-token
+        uses: actions/create-github-app-token@v2
+        with:
+          app-id: ${{ secrets.MINDHIVE_CICD_GH_APP_ID }}
+          private-key: ${{ secrets.MINDHIVE_CICD_GH_PRIVATE_KEY }}
+          owner: mindhivenz
+          # App must be installed on both repos:
+          #  - deploy-js: contents:read
+          #  - deploy-py: contents:write, pull_requests:write
+          repositories: |
+            deploy-js
+            deploy-py
+
+      - name: Checkout deploy-js (source of truth)
+        uses: actions/checkout@v4
+        with:
+          token: ${{ steps.app-token.outputs.token }}
+          fetch-depth: 0
+
+      - name: Port changes to deploy-py and open PR
+        uses: anthropics/claude-code-action@v1
+        env:
+          GH_TOKEN: ${{ steps.app-token.outputs.token }}
+        with:
+          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          github_token: ${{ steps.app-token.outputs.token }}
+          claude_args: |
+            --model claude-opus-4-8
+            --max-turns 40
+            --allowedTools Edit,Read,Write,Glob,Grep,Bash
+          prompt: |
+            Use the update-deploy-py skill to port any changes in this repo
+            (deploy-js, the source of truth) into the deploy-py Python port and
+            open a PR on deploy-py if it has drifted. Follow the skill exactly.

script/mhd

@@ -1,32 +1,43 @@
 #!/usr/bin/env bash
-# Find project deploy directory and run gulp tasks in there
-# Extra args are packaged into an environment var to be parsed by src/internal/args.ts
+# Find project deploy directory and run deploy tasks in there
+# Extra args are packaged into MHD_ARGS env var to be parsed by the task runner
 
 set -e
 ancestor_dir=$(git rev-parse --show-toplevel)
 project_dir=${ancestor_dir%%-control}
 cd "${project_dir}/deploy" >/dev/null
+
 vault_opts=""
 mhd_args=""
+task_args=""
 if [[ $# == 0 ]]; then
-  gulp_args="--tasks  --compact-tasks"
+  task_args=""
 elif [[ $1 =~ ^--.* ]]; then
-  gulp_args="$*"
+  task_args="$*"
 else
-  gulp_args="$1"
+  task_args="$1"
   shift
   IFS=$'\t'
   mhd_args="$*"
 fi
-if [[ $gulp_args =~ ^open:.* ]]; then
+if [[ $task_args =~ ^open:.* ]]; then
   vault_opts="--duration=1h"
 fi
 
+if [[ -f gulpfile.ts ]]; then
+  # TypeScript deploy: gulp tasks in gulpfile.ts
+  [[ -z $task_args ]] && task_args="--tasks --compact-tasks"
+  runner=(env TS_NODE_TRANSPILE_ONLY=true node_modules/.bin/gulp)
+else
+  # Python deploy: uv + typer tasks in tasks.py
+  [[ -z $task_args ]] && task_args="--help"
+  runner=(uv run tasks.py)
+fi
+
 if [ "$CI" = "true" ]; then
   # shellcheck disable=SC2086
-  TS_NODE_TRANSPILE_ONLY=true MHD_ARGS="$mhd_args" node_modules/.bin/gulp $gulp_args
+  MHD_ARGS="$mhd_args" "${runner[@]}" $task_args
 else
   # shellcheck disable=SC2086
-  TS_NODE_TRANSPILE_ONLY=true MHD_ARGS="$mhd_args" aws-vault exec $vault_opts mindhive-ops -- \
-    node_modules/.bin/gulp $gulp_args
+  MHD_ARGS="$mhd_args" aws-vault exec $vault_opts mindhive-ops -- "${runner[@]}" $task_args
 fi