Add end-to-end install integration test harness

Agent-Logs-Url: https://github.com/githubnext/autoloop/sessions/df881075-c8de-46ce-8314-6945f02f4f05

Co-authored-by: mrjf <180956+mrjf@users.noreply.github.com>

Author: Copilot

.github/workflows/install-integration-test.yml

@@ -0,0 +1,54 @@
+name: Install Integration Test
+
+# End-to-end test of install.md against a long-lived target repo.
+# Manual-dispatch only -- this exercises real LLM calls and force-pushes a
+# remote branch, so it must not run on PRs or schedules.
+
+on:
+  workflow_dispatch:
+    inputs:
+      keep_state_on_failure:
+        description: "Leave test repo in failure state for inspection"
+        type: boolean
+        default: false
+      install_test_repo:
+        description: "Target repo for the install (owner/repo)"
+        type: string
+        default: "mrjf/autoloop-test"
+
+jobs:
+  install-integration:
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install gh aw extension
+        env:
+          GH_TOKEN: ${{ secrets.INSTALL_TEST_TOKEN }}
+        run: gh extension install github/gh-aw
+
+      - name: Install Copilot CLI
+        env:
+          GH_TOKEN: ${{ secrets.INSTALL_TEST_TOKEN }}
+        # The Copilot CLI is distributed as an npm package. If the install
+        # path changes upstream, update this single step.
+        run: npm install -g @github/copilot
+
+      - name: Verify gh auth
+        env:
+          GH_TOKEN: ${{ secrets.INSTALL_TEST_TOKEN }}
+        run: gh auth status
+
+      - name: Run integration test
+        env:
+          GH_TOKEN: ${{ secrets.INSTALL_TEST_TOKEN }}
+          INSTALL_TEST_REPO: ${{ inputs.install_test_repo }}
+          KEEP_STATE_ON_FAILURE: ${{ inputs.keep_state_on_failure && '1' || '0' }}
+        run: ./tests/install-integration/run.sh

tests/install-integration/README.md

@@ -0,0 +1,67 @@
+# install-integration
+
+End-to-end integration test for [`install.md`](../../install.md). Runs the
+install flow as a real coding agent (Copilot CLI) against a long-lived
+target repo (`mrjf/autoloop-test` by default), then exercises Phase 2 by
+running one iteration each of three programs across the program-source ×
+strategy matrix.
+
+This test is **manual-dispatch only**. It is not part of CI.
+
+## Local mode
+
+```bash
+# from the autoloop repo root:
+./tests/install-integration/run.sh
+```
+
+Requirements:
+
+- `gh` CLI authenticated as a user with write access to the target repo.
+- `copilot` CLI on PATH.
+- `python3` and `git` on PATH.
+
+Optional env / flags:
+
+- `INSTALL_TEST_REPO=<owner>/<repo>` -- override the target (default
+  `mrjf/autoloop-test`).
+- `--keep` (or `KEEP_STATE_ON_FAILURE=1`) -- skip teardown on failure so
+  the failure state can be inspected. Run `teardown.sh <repo> <base-sha>`
+  manually afterwards.
+
+## Actions mode
+
+Trigger the **Install Integration Test** workflow from the Actions tab. It
+runs the same script on a GitHub-hosted runner. Requires the
+`INSTALL_TEST_TOKEN` repo secret -- a PAT with `repo` scope on the target
+repo (the default `GITHUB_TOKEN` has no access to repos outside the host).
+
+## What it tests
+
+See [the issue that introduced this harness](https://github.com/githubnext/autoloop/issues)
+for the full motivation. In short:
+
+- **Phase 1** (file presence + lock idempotency) -- catches regressions in
+  `install.md` and in `gh aw compile`.
+- **Phase 2** (3 programs × 1 iteration each) -- catches regressions in
+  the scheduler, in strategy discovery, and in the iteration loop. The
+  three programs cover:
+
+  | # | Source     | Strategy        |
+  |---|------------|-----------------|
+  | 1 | file-based | OpenEvolve      |
+  | 2 | issue-based| Test-Driven     |
+  | 3 | file-based | plain (default) |
+
+- **Phase 3** (teardown) -- resets the target repo to the captured base
+  SHA, closes test issues/PRs, and deletes test branches.
+
+## Files
+
+| File                  | Purpose                                          |
+|-----------------------|--------------------------------------------------|
+| `run.sh`              | Driver. Orchestrates phases 1-3.                 |
+| `prompt.md`           | Prompt fed to Copilot CLI (edit without touching the driver). |
+| `verify-phase1.sh`    | File-presence + lock-idempotency assertions.     |
+| `verify-phase2.sh`    | Per-program assertions (one call per program).   |
+| `teardown.sh`         | Idempotent cleanup. Safe to re-run.              |

tests/install-integration/prompt.md

@@ -0,0 +1,17 @@
+You are installing autoloop into a freshly-reset GitHub repository.
+
+Your working directory is the root of that repository, cloned locally. The
+repository is empty except for the base fixtures in `src/` and `tests/`.
+
+Follow the install instructions at the URL below, EXACTLY AS WRITTEN. Execute
+each step using shell commands. Do not skip steps. Do not improvise. Do not
+optimize or "improve" the instructions.
+
+Stop after Step 5 (the install PR is opened). Do NOT proceed to Step 6
+("Create Your First Program") -- the test harness handles program creation
+itself in a deterministic way.
+
+When you finish: print a single line `INSTALL_PR=<url>` with the URL of the
+PR you opened in step 5. Then stop.
+
+Install instructions: https://github.com/githubnext/autoloop/blob/main/install.md