Add the Gitea Actions configuration

The third port of the same two things. Gitea walks WORKFLOW_DIRS and stops
at the first directory that exists, so .gitea/workflows/ replaces
.github/workflows/ outright on a Gitea instance — which is the point, since
deploy.yml there is built on repository_dispatch and an environment: and
Gitea has neither.

Four differences shape the files:

- No environment:, so no scoped secret and no required reviewers. The file
  does not write the key at all rather than claim a protection that Gitea
  parses and ignores; what gates a deploy is write access to the repository.
- No repository_dispatch. External callers post to the workflow dispatch
  API, which fills in the same form — and can return a run id, so unlike
  repository_dispatch the caller may follow the deploy it asked for.
- runs-on takes a literal label only, so GitHub's vars.DEPLOY_RUNNER
  expression becomes one documented line to edit.
- A job is itself a container with no Docker socket, so the secret scan runs
  the pinned gitleaks binary instead of the upstream image.

Input validation carries over unchanged and matters more here: type: choice
constrains the dispatch form, not the API.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>

.gitea/workflows/connectivity.yml

@@ -0,0 +1,131 @@
+---
+# Diagnostics, for when a deploy job hangs instead of failing.
+#
+# Two jobs, deliberately separate:
+#
+#   ping   uses the real deploy key and asks Ansible to talk to every host. This
+#          is the right first thing to run after setting the secret up, and the
+#          right first thing to run when a deploy misbehaves. It changes nothing.
+#   probe  uses no credentials at all and answers a narrower question: where do
+#          the packets stop? Run it when ping times out.
+#
+# The distinction that matters, and the reason `probe` exists:
+#
+#   Permission denied (publickey)  the key is wrong — you reached sshd
+#   Connection refused             you reached the host, nothing is listening
+#   Connection timed out           packets are being dropped in transit
+#
+# Only the third is a network problem, and no amount of re-pasting the key will
+# fix it. See docs/connectivity.md.
+name: connectivity
+
+on:
+  workflow_dispatch:
+    inputs:
+      hosts:
+        description: Space-separated hosts to probe (defaults to the inventory)
+        required: false
+        default: edge.example.com app.example.com
+
+jobs:
+  ping:
+    name: ansible ping
+    # A literal label, for the reason spelled out in deploy.yml. Keep the two
+    # files in step: a diagnostic that runs somewhere other than the deploy is
+    # diagnosing the wrong machine.
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - name: Install the pinned toolchain
+        run: |
+          set -eu
+          scripts/check.sh --print-specs | head -1 | xargs -d '\n' pip install --quiet
+      - name: Load the deploy key
+        env:
+          SSH_KEY_BACKEND: ci
+          SSH_PRIVATE_KEY_B64: ${{ secrets.SSH_PRIVATE_KEY_B64 }}
+        run: |
+          set -eu
+          SSH_KEY_PATH="$(scripts/load-ssh-key.sh)"
+          echo "SSH_KEY_PATH=$SSH_KEY_PATH" >> "$GITHUB_ENV"
+      # Success looks like `SUCCESS => {"ping": "pong"}` for every host. That
+      # proves the runner decoded the key, reached port 22, authenticated, and ran
+      # Python on the far side — which is everything a deploy needs except the
+      # playbook itself.
+      - name: Ping every host in the inventory
+        working-directory: ansible
+        run: ansible all -m ansible.builtin.ping
+
+  probe:
+    name: network probe
+    runs-on: ubuntu-latest
+    # A report, not a gate. Every section must run even when the thing it probes
+    # is unreachable, which is the normal case when you are running this. The
+    # steps also keep themselves green with `set +e`, so the report survives on an
+    # instance that ignores this key.
+    continue-on-error: true
+    steps:
+      - name: Probe
+        env:
+          HOSTS: ${{ github.event.inputs.hosts }}
+        run: |
+          set +e
+
+          # A bare TCP connect can be answered by a transparent proxy that then
+          # says nothing, which looks like success and is not. Real sshd greets
+          # first, so the banner is the only honest proof you reached the server.
+          ssh_banner() {
+            b=$(timeout 6 bash -c "exec 3<>/dev/tcp/$1/${2:-22}; head -c 12 <&3" 2>/dev/null || true)
+            case "$b" in
+              SSH-*) printf '  %-26s %-5s OK, sshd replied: %s\n' "$1" "${2:-22}" "$b" ;;
+              '')    printf '  %-26s %-5s NO REPLY (dropped, or a silent proxy)\n' "$1" "${2:-22}" ;;
+              *)     printf '  %-26s %-5s answered, but not sshd: %s\n' "$1" "${2:-22}" "$b" ;;
+            esac
+          }
+
+          # Distinguishes "refused" (allowed out, nothing listening) from
+          # "filtered" (dropped in transit). This is the whole point of the job:
+          # if every port but 443 is filtered, you are behind an egress allowlist
+          # and no alternative SSH port will help.
+          port_state() {
+            # `rc=0; out=$(...) || rc=$?` and not `out=$(...); rc=$?` — the latter
+            # is a failing simple command, which under set -e ends the step before
+            # rc is ever read, so a filtered port aborts the report instead of
+            # describing it.
+            rc=0
+            out=$(timeout 5 bash -c "cat < /dev/null > /dev/tcp/$1/$2" 2>&1) || rc=$?
+            if [ "$rc" -eq 0 ]; then                          s='open'
+            elif printf '%s' "$out" | grep -qi refused; then
+              s='refused -> port allowed out'
+            elif [ "$rc" -eq 124 ]; then
+              s='filtered -> blocked in transit'
+            else
+              s="unclear (rc=$rc) $out"
+            fi
+            printf '  %-26s %-5s %s\n' "$1" "$2" "$s"
+          }
+
+          printf '== 0. does this runner have a network at all? ==\n'
+          printf '  DNS:            '; getent hosts github.com >/dev/null 2>&1 && echo resolves || echo 'CANNOT RESOLVE'
+          # Egress to github.com is worth knowing for a second reason here: unless
+          # your administrator set DEFAULT_ACTIONS_URL to `self`, it is also where
+          # every `uses:` step is downloaded from.
+          printf '  https external: '
+          curl -s -o /dev/null -w '%{http_code}\n' --max-time 10 \
+            https://github.com || echo 'NO EGRESS'
+          printf '  egress address: '; curl -s --max-time 10 https://ifconfig.me || echo unknown
+
+          printf '\n\n== 1. SSH to the target hosts ==\n'
+          for h in $HOSTS; do ssh_banner "$h"; done
+
+          printf '\n== 2. SSH elsewhere: is outbound 22 blocked wholesale? ==\n'
+          ssh_banner github.com
+
+          printf '\n== 3. Which ports are allowed out at all? ==\n'
+          for h in $HOSTS; do
+            port_state "$h" 22
+            port_state "$h" 443
+          done

.gitea/workflows/deploy.yml

@@ -0,0 +1,167 @@
+---
+# Deploys, on Gitea. One way in, and only one:
+#
+#   - a human fills in the form under Actions > deploy > Run workflow, or the
+#     same form is filled in over the API by another repository's job
+#
+# Nothing deploys on a push. That is the single most important property of this
+# file: a merge to the default branch changes what *would* be deployed, and a
+# person still decides when.
+#
+# Gitea has no `repository_dispatch`, so the two entry points GitHub gets collapse
+# into one: the workflow dispatch API posts to *this* form.
+#
+#   POST /api/v1/repos/<owner>/<repo>/actions/workflows/deploy.yml/dispatches
+#
+# That is better than it sounds — the response can carry the run id, so a caller
+# can follow the deploy it asked for, which `repository_dispatch` never allowed.
+# See docs/triggering-deploys.md.
+#
+# The `options:` list has to be kept in step with ansible/playbooks/ by hand; a
+# choice input cannot be populated dynamically on any platform. It is step 4 of
+# skills/adding-a-stack/SKILL.md. Setup: docs/ci-gitea.md.
+name: deploy
+
+on:
+  workflow_dispatch:
+    inputs:
+      stack:
+        description: Which stack to act on
+        required: true
+        type: choice
+        options:
+          - site
+          - reverse-proxy
+          - banner
+          - metrics
+          - static-site
+          - webapp
+          - webapp-staging
+      mode:
+        description: check reports what would change; deploy does it
+        required: true
+        default: check
+        type: choice
+        options:
+          - check
+          - deploy
+      pull:
+        description: >-
+          Image pull policy. always fetches newer images and enables per-stack
+          update work (backups, migrations).
+        required: true
+        default: policy
+        type: choice
+        options:
+          - policy
+          - always
+          - missing
+          - never
+
+# Two deploys of the same stack at once would race on the host. Concurrency
+# syntax landed in Gitea 1.26; before that this block is ignored, and the only
+# thing standing between you and a race is that both runs are started by hand.
+concurrency:
+  group: deploy-${{ github.event.inputs.stack }}
+  cancel-in-progress: false
+
+jobs:
+  ansible:
+    name: ${{ github.event.inputs.mode || 'check' }} ${{ github.event.inputs.stack }}
+    # Change this label if the runner that picks up `ubuntu-latest` cannot reach
+    # your hosts, and register a second runner carrying whatever you put here.
+    # It has to be a literal: Gitea supports `runs-on: xyz` and `runs-on: [xyz]`
+    # only, so GitHub's `${{ vars.DEPLOY_RUNNER || 'ubuntu-latest' }}` trick has
+    # no equivalent. Do not name a label no runner carries — the job then queues
+    # forever rather than failing.
+    runs-on: ubuntu-latest
+    # No `environment:` here. Gitea parses and ignores it, so writing one would
+    # claim a protection that does not exist: there are no required reviewers and
+    # no environment-scoped secrets. SSH_PRIVATE_KEY_B64 is an ordinary
+    # repository (or organisation) secret, and what actually gates a deploy is
+    # that dispatching a workflow needs write access to this repository. See
+    # "What replaces environments" in docs/ci-gitea.md.
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      # STACK, MODE and PULL reach a shell command, and over the dispatch API they
+      # arrive from another repository. They go through the environment, never
+      # through ${{ }} interpolation into a run: block — an expression is
+      # substituted into the script *before* the shell sees it, so a crafted value
+      # would be executed rather than compared. Validate, then use.
+      #
+      # `type: choice` is a UI constraint, not an API one. Do not read the
+      # dropdown above as validation.
+      - name: Validate the requested stack and pull policy
+        env:
+          # The `|| default` halves are not decoration. An input's declared
+          # `default:` is applied by the dispatch *form*; a caller posting to the
+          # API can leave the key out altogether, and a missing `mode` would then
+          # be refused below rather than falling back to the safe value. `stack`
+          # has no default on purpose — there is no safe guess for it.
+          STACK: ${{ github.event.inputs.stack }}
+          MODE: ${{ github.event.inputs.mode || 'check' }}
+          PULL: ${{ github.event.inputs.pull || 'policy' }}
+        run: |
+          set -eu
+          case "$STACK" in
+            ''|*[!a-z0-9_-]*) echo "refusing STACK='$STACK'"; exit 1 ;;
+          esac
+          case "$MODE" in
+            check|deploy) ;;
+            *) echo "refusing MODE='$MODE'"; exit 1 ;;
+          esac
+          # An unchecked value can smuggle extra `-e` overrides onto the command
+          # line (stack_dest, ansible_host) and redirect the deploy somewhere else
+          # entirely.
+          case "$PULL" in
+            policy|always|missing|never) ;;
+            *) echo "refusing PULL='$PULL'"; exit 1 ;;
+          esac
+          test -f "ansible/playbooks/${STACK}.yml" \
+            || { echo "no playbook for '$STACK'"; exit 1; }
+          {
+            echo "STACK=$STACK"
+            echo "MODE=$MODE"
+            echo "PULL=$PULL"
+          } >> "$GITHUB_ENV"
+
+      - name: Install the pinned toolchain
+        run: |
+          set -eu
+          scripts/check.sh --print-specs | head -1 | xargs -d '\n' pip install --quiet
+          ansible-galaxy install -r ansible/requirements.yml
+
+      # Two lines, not `export SSH_KEY_PATH="$(...)"`. export is a special builtin,
+      # so the compound command reports *its* exit status and swallows the
+      # script's — under `set -e` a loader failure goes unnoticed and the job dies
+      # much later with a per-host "Permission denied (publickey)" instead of the
+      # loader's own message saying which variable was missing.
+      - name: Load the deploy key
+        env:
+          SSH_KEY_BACKEND: ci
+          SSH_PRIVATE_KEY_B64: ${{ secrets.SSH_PRIVATE_KEY_B64 }}
+        run: |
+          set -eu
+          SSH_KEY_PATH="$(scripts/load-ssh-key.sh)"
+          echo "SSH_KEY_PATH=$SSH_KEY_PATH" >> "$GITHUB_ENV"
+
+      - name: Run the playbook
+        working-directory: ansible
+        env:
+          ANSIBLE_FORCE_COLOR: "1"
+        run: |
+          set -eu
+          # An `if`, not `[ ... ] && extra=...`: when the test fails the AND-list
+          # returns non-zero, and under `set -e` that ends the job right here with
+          # no error message and a green-looking cancel.
+          if [ "$MODE" = "check" ]; then
+            set -- --check --diff
+          else
+            set --
+          fi
+          ansible-playbook "playbooks/${STACK}.yml" "$@" -e "pull=$PULL"

.gitea/workflows/lint.yml

@@ -0,0 +1,112 @@
+---
+# The static gate on Gitea. The same three checks as .github/workflows/lint.yml,
+# with the adjustments a Gitea runner forces. Setup: docs/ci-gitea.md.
+#
+# Creating .gitea/workflows/ at all is a decision rather than a convenience.
+# Gitea walks WORKFLOW_DIRS in order — `.gitea/workflows` then `.github/workflows`
+# by default — and stops at the first directory that exists. So this directory
+# replaces the GitHub workflows wholesale on a Gitea instance, which is what you
+# want: deploy.yml there is built on `repository_dispatch` and an `environment:`,
+# and Gitea has neither.
+#
+# Delete whichever platform you are not using rather than letting it rot.
+name: lint
+
+on:
+  push:
+    branches: ["**"]
+  pull_request:
+
+# Concurrency syntax landed in Gitea 1.26. Older instances ignore this block
+# rather than failing on it — you simply get both runs.
+concurrency:
+  group: lint-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  ansible:
+    name: yamllint + ansible-lint + syntax
+    # A literal label, never `${{ vars.SOMETHING }}`: Gitea supports `runs-on: xyz`
+    # and `runs-on: [xyz]` and nothing more. There are also no hosted runners, so
+    # this label is only ever your own — see "Runners" in docs/ci-gitea.md.
+    runs-on: ubuntu-latest
+    steps:
+      # Resolved against DEFAULT_ACTIONS_URL, which is github.com unless your
+      # administrator set it to `self`. On an instance that cannot reach github.com
+      # this step is the first thing that fails, and it fails for every workflow —
+      # mirror actions/checkout into the instance, or set the setting back.
+      - uses: actions/checkout@v4
+
+      # Keep this even though the runner image ships a Python. Ubuntu 24.04 images
+      # are PEP 668 "externally managed", where a plain `pip install` refuses to
+      # touch the system interpreter; setup-python brings its own and sidesteps it.
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      # scripts/check.sh is the single source of the version pins; it prints them
+      # so that this workflow installs exactly what a contributor runs locally.
+      # Nothing here can drift from `make check`.
+      - name: Install the pinned toolchain
+        run: |
+          set -eu
+          scripts/check.sh --print-specs | tee /dev/stderr | xargs -d '\n' pip install --quiet
+
+      - name: Run the static gate
+        run: RUNNER=installed scripts/check.sh
+
+  shell:
+    name: shellcheck
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      # Jobs run as root inside the runner's container, where `sudo` is often not
+      # installed at all. Asking for it only when the uid is not 0 keeps this step
+      # working on a container runner and on a `:host` runner alike.
+      - name: Install shellcheck
+        run: |
+          set -eu
+          if ! command -v shellcheck >/dev/null 2>&1; then
+            sudo=""
+            [ "$(id -u)" -eq 0 ] || sudo=sudo
+            $sudo apt-get update -qq
+            $sudo apt-get install -y --no-install-recommends shellcheck
+          fi
+      - name: Check the shell scripts
+        run: shellcheck scripts/*.sh server/*/*/app/*.sh
+
+  secrets:
+    name: secret scan
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          # detect scans history, not just the working tree. A secret that was
+          # committed and then removed is still a leaked secret.
+          fetch-depth: 0
+
+      # The GitHub workflow runs the upstream gitleaks *image*. That is not
+      # portable here: an act_runner job is itself a container, and it has no
+      # Docker socket unless the administrator mounted one into it, so
+      # `docker run` fails with "Cannot connect to the Docker daemon". The
+      # released binary needs nothing but curl and has the same behaviour.
+      #
+      # Pinned deliberately: a scanner that changes under you turns a green
+      # history red on a morning you changed nothing.
+      - name: Fetch the pinned gitleaks binary
+        env:
+          GITLEAKS_VERSION: 8.30.1
+        run: |
+          set -eu
+          base=https://github.com/gitleaks/gitleaks/releases/download
+          curl -sSfL -o /tmp/gitleaks.tar.gz \
+            "$base/v${GITLEAKS_VERSION}/gitleaks_${GITLEAKS_VERSION}_linux_x64.tar.gz"
+          # Left in /tmp and run from there. Installing into /usr/local/bin would
+          # need root, which a `:host` runner does not have and should not want.
+          tar -xzf /tmp/gitleaks.tar.gz -C /tmp gitleaks
+          chmod +x /tmp/gitleaks
+
+      - name: Scan the working tree and history for committed secrets
+        run: |
+          /tmp/gitleaks detect --source=. --config=.gitleaks.toml \
+            --redact --verbose --no-banner

docs/ci-gitea.md

@@ -0,0 +1,225 @@
+# CI on Gitea
+
+Three workflows, in `.gitea/workflows/`:
+
+| Workflow | Trigger | Needs host access |
+| --- | --- | --- |
+| [`lint.yml`](../.gitea/workflows/lint.yml) | every push and pull request | no |
+| [`deploy.yml`](../.gitea/workflows/deploy.yml) | manual, or the dispatch API | yes |
+| [`connectivity.yml`](../.gitea/workflows/connectivity.yml) | manual | yes (`ping` only) |
+
+Gitea Actions is close enough to GitHub Actions that the files look almost the
+same, and different enough in four specific places that copying the GitHub ones
+across produces a repository that lints fine and cannot deploy. Those four
+places are the whole content of this document.
+
+**Version floor.** `workflow_dispatch` arrived in Gitea **1.23**, dispatching it
+over the API in **1.24**, and `concurrency:` in **1.26**. On 1.23 the deploy
+form works and only external triggering is unavailable; below 1.23 nothing here
+runs manually at all.
+
+## `.gitea/workflows/` shadows `.github/workflows/`
+
+Gitea walks `WORKFLOW_DIRS` — `.gitea/workflows` then `.github/workflows` by
+default — and stops at **the first directory that exists**. So adding this
+directory silently switches a Gitea instance over to these three files and
+leaves the GitHub ones unread.
+
+That is the intent, not a side effect. `.github/workflows/deploy.yml` is built
+on `repository_dispatch` and an `environment:`, and Gitea has neither: on Gitea
+it would be a deploy workflow with one entry point missing and a protection that
+looks configured and is not.
+
+If you are only using one platform, delete the others rather than letting them
+rot. Both ports are kept here because the translation between them is a large
+part of what this repository is demonstrating.
+
+## Setup
+
+### 1. A runner
+
+There are no hosted runners. Nothing in Actions runs until you register an
+`act_runner` yourself, so this step is not optional the way it is on GitHub.
+
+```bash
+act_runner register --no-interactive \
+  --instance https://gitea.example.com \
+  --token <registration token from Settings → Actions → Runners> \
+  --labels ubuntu-latest:docker://docker.gitea.com/runner-images:ubuntu-latest
+```
+
+A label is `<name>[:<schema>[:<args>]]`, where the schema is `docker://<image>`
+(the job runs in a container from that image) or `host` (the job runs directly
+on the machine with the tools installed there). `runs-on` is matched against the
+label *names*, first match wins, and a job whose `runs-on` matches nothing falls
+back to `docker.gitea.com/runner-images:ubuntu-latest`.
+
+**Use the full image, not a `-slim` one.** The slim images are `node:20-slim`
+with very little else. Every `uses:` step is a JavaScript action, so the image
+must have Node — that much the slim images give you — but the rest of these jobs
+want `curl`, `git`, `tar` and a working `apt-get`.
+
+**If the runner cannot reach your hosts**, register a second one somewhere that
+can, give it a label of its own, and change `runs-on:` in `deploy.yml` and in
+`connectivity.yml`. It has to be a literal — Gitea supports `runs-on: xyz` and
+`runs-on: [xyz]` and nothing else, so GitHub's
+`${{ vars.DEPLOY_RUNNER || 'ubuntu-latest' }}` has no equivalent here. Do not
+name a label before a runner carries it; the job then queues forever rather than
+failing.
+
+A runner only makes *outbound* HTTPS connections to the instance, which is why
+this works where an inbound firewall exception would need a policy change.
+
+### 2. The deploy key
+
+Settings → Actions → Secrets → Add Secret, named `SSH_PRIVATE_KEY_B64`, holding
+the base64 of your deploy private key. [secrets.md](secrets.md) has how to
+encode it, and what not to use on Windows.
+
+An organisation-level secret of the same name works too, and is the right choice
+when several config repositories share one deploy key. A repository secret wins
+over an organisation one of the same name.
+
+The matching public key must be in the deploy user's `authorized_keys` on every
+host.
+
+### 3. What replaces environments
+
+Nothing does, and it is worth being clear-eyed about that before you point this
+at production.
+
+`jobs.<job_id>.environment` is **parsed and ignored** by Gitea. There are no
+environment-scoped secrets and no required reviewers, so the deploy key is
+readable by any workflow run in this repository, and "press the button" stays a
+one-person operation. `deploy.yml` therefore does not write an `environment:`
+key at all — a protection that exists only in YAML is worse than no protection,
+because you stop checking.
+
+What you actually have:
+
+- **Dispatching a workflow requires write access** to the repository. That is
+  the real gate on who can deploy, so the repository's collaborator list *is*
+  your deploy ACL. Keep it short.
+- **Protected branches** stop a deploy being dispatched from a ref nobody
+  reviewed — the dispatch form takes a branch, and `ansible/` is read from the
+  ref you pick. Protect the default branch and restrict who may push to it.
+- **An organisation secret with limited membership**, if you want the key
+  scoped more tightly than a repository full of contributors.
+
+If you need two-person deploys, the honest options are a Gitea instance running
+a merge-approval workflow in front of the branch the deploy runs from, or
+deploying from a repository only the deployers can write to.
+
+## Deploying
+
+Actions → **deploy** → Run workflow. Three fields:
+
+| Field | Values | Meaning |
+| --- | --- | --- |
+| `stack` | the dropdown | Which playbook to run. `site` is all of them. |
+| `mode` | `check`, `deploy` | `check` runs `--check --diff` and touches nothing |
+| `pull` | `policy`, `always`, `missing`, `never` | `always` fetches newer images **and** enables per-stack update work |
+
+Work up from harmless to real the first time:
+
+| Step | What to run | What it proves |
+| --- | --- | --- |
+| 1 | **connectivity** → `ping` | The runner can decode the key, reach every host over SSH, and run Python there. Changes nothing. |
+| 2 | **deploy** with `mode: check` | What that deploy *would* change. |
+| 3 | **deploy** with `mode: deploy` | The real thing. |
+
+`static-site` is the safest stack to prove this with: one container behind the
+proxy, no database and no persistent state, so redeploying it costs nothing if
+it goes wrong.
+
+Another repository can ask this one to deploy after building an image — see
+[triggering-deploys.md](triggering-deploys.md). On Gitea that is the same form,
+filled in over the API.
+
+## Differences from GitHub Actions that shape these files
+
+| | Gitea | GitHub Actions |
+| --- | --- | --- |
+| Manual run | `workflow_dispatch` form (1.23+) | `workflow_dispatch` form |
+| External trigger | `POST …/actions/workflows/deploy.yml/dispatches` (1.24+) | `repository_dispatch` |
+| Caller waits for the result | Can poll: the dispatch response carries the run id | Not possible |
+| Restricting who can deploy | Write access + protected branches | Environment protection rules |
+| Runner selection | A literal label in the file | `runs-on:` with a `vars.` expression |
+| Hosted runners | None; you register `act_runner` | Ubuntu, macOS, Windows |
+| `concurrency:` | 1.26+, ignored before | Supported |
+| Secret scan | The pinned gitleaks binary | The upstream gitleaks Docker image |
+
+**Expressions are barely supported.** Of GitHub's expression functions only
+`always()` is implemented, so `success()`, `failure()`, `contains()`,
+`fromJSON()` and friends are not available. The operators (`||`, `&&`, `==`) and
+the contexts (`github`, `secrets`, `vars`, `env`) work, which is all these three
+files use. Keep it that way if you restructure them.
+
+**`docker run` does not work in a job.** An `act_runner` job is itself a
+container and has no Docker socket unless the administrator mounted one in, so
+the GitHub lint workflow's `docker run … gitleaks` fails with "Cannot connect to
+the Docker daemon". The released binary is fetched and run directly instead —
+pinned, because a scanner that changes under you turns a green history red on a
+morning you changed nothing.
+
+**`actions/setup-python` stays, for a Gitea-specific reason.** The runner images
+are Ubuntu 24.04, where the system Python is PEP 668 "externally managed" and a
+plain `pip install` refuses to touch it. `setup-python` brings its own
+interpreter and sidesteps that. If you drop it, you need
+`pip install --break-system-packages` and you are then installing into the image.
+
+**Every `uses:` is downloaded from github.com** unless your administrator set
+`DEFAULT_ACTIONS_URL = self`, in which case `actions/checkout` and
+`actions/setup-python` must be mirrored into the instance as repositories of
+those names, or every workflow here fails on its first step. That setting takes
+only `github` or `self`; an arbitrary mirror URL has to be spelled out in full
+in the workflow.
+
+## Why untrusted input is validated
+
+`deploy.yml` matches `stack` against `[a-z0-9_-]+` and `pull` against a fixed
+list before running anything, and confirms the playbook exists. With the values
+coming from a dropdown this looks redundant twice over.
+
+It is not, for two reasons specific to this platform. `type: choice` is a
+constraint on the *form*, not on the API: a dispatch call may send any string
+for `stack`, and the dispatch API is the only way another repository can trigger
+a deploy here, so every externally triggered deploy arrives through exactly the
+path the dropdown does not cover.
+
+And the values never go into a `run:` block through `${{ }}`. A workflow
+expression is substituted into the script text *before the shell sees it*, so a
+value containing shell syntax is executed rather than compared. Values go
+through `env:` and are read as `"$STACK"`:
+
+```yaml
+        env:
+          STACK: ${{ github.event.inputs.stack }}
+        run: |
+          case "$STACK" in
+            ''|*[!a-z0-9_-]*) echo "refusing STACK='$STACK'"; exit 1 ;;
+          esac
+```
+
+Keep that shape if you restructure the job. It is the difference between a
+validated argument and remote code execution on your deploy runner.
+
+## Troubleshooting
+
+| Job fails with | Cause |
+| --- | --- |
+| Nothing runs at all; no run appears | Actions is disabled for the repository (Settings → Repository → Enable Actions), or no runner is registered |
+| The workflows in `.github/` run instead | `.gitea/workflows/` is missing from the ref you dispatched, or `WORKFLOW_DIRS` was reordered instance-wide |
+| No **Run workflow** button | Gitea older than 1.23, or you do not have write access to the repository |
+| `exec: "node": executable file not found` | The runner's image has no Node, so no `uses:` step can run. Use a `runner-images` tag rather than a bare `alpine`/`python` image |
+| `Unable to resolve action actions/checkout` | The instance cannot reach github.com, or `DEFAULT_ACTIONS_URL = self` and the action is not mirrored |
+| `externally-managed-environment` from pip | The `actions/setup-python` step was removed |
+| `Cannot connect to the Docker daemon` | A step is trying to run Docker inside the job container |
+| Job queued forever | `runs-on:` names a label no runner carries |
+| `set SSH_PRIVATE_KEY_B64, or SSH_PRIVATE_KEY …` | The secret is not on this repository or its organisation |
+| `does not look like a private key` | The value was encoded with `certutil` or `Get-Content`, or re-wrapped across lines |
+| `The role 'compose_stack' was not found` | Ansible refused a world-writable working directory. Set `ANSIBLE_CONFIG=$GITHUB_WORKSPACE/ansible/ansible.cfg` on the job |
+| `Permission denied (publickey)` | The matching public key is not in the deploy user's `authorized_keys` |
+| `Connection timed out` | Network path, not credentials — run **connectivity** → `probe`, and read [connectivity.md](connectivity.md) |
+| `no playbook for '…'` | The `stack` value does not match a file in `ansible/playbooks/` |
+| Deploy succeeds, nothing changed | Images are tagged `latest` and `pull` was `policy` |