feat(templates): add a fixture stack and platform the per-stack checks read (#1925) (#1948)

stack-fixture and platform-fixture live under templates/fixture/ and carry
fixture: true in the manifest. resolve.py's listings and cache and every
block sync.py writes leave them out; a fixture still resolves when named.

The smoke checks that read every stack now select by the manifest's
stacks: key instead of the templates/stack/ directory, so they keep an
input once the concrete stacks move to forks. SYS-14's cases name the
fixtures, tests/context-tiers.txt records the fixture category at 200K,
and the e2e canary moves from STK-15 to a new STK-24 on the fixture stack.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

Author: braboj

CHANGELOG.md

@@ -21,6 +21,13 @@ alters no template carries no entry.
 
 ## [Unreleased]
 
+### Added
+
+- `templates/fixture/` adds `stack-fixture` and `platform-fixture`, test
+  input marked `fixture: true` in the manifest. Listings and `generated/`
+  omit them; per-stack checks select by the manifest's `stacks:` key, and
+  the e2e canary is now STK-24 on the fixture.
+
 ### Breaking changes
 
 - The concern directories leave `templates/base/`: `templates/base/<concern>/`

CLAUDE.md

@@ -29,6 +29,7 @@ templates/          # All template source files
   backend/          # Backend layer — HTTP, API, database, observability
   frontend/         # Frontend layer — UX, accessibility, CSS, SSG
   platform/         # CI and security tool mappings per hosting platform
+  fixture/          # Test-input stack and platform — checked, never listed
   stack/            # Concrete stacks — extend base + layer templates
   INTERVIEW.md      # Agent-driven project setup interview
   manifest.yaml     # Machine-readable dependency graph
@@ -364,7 +365,7 @@ py tests/run_smoke.py SYS-01       # run one check by ID
 py tests/run_conformance.py        # the templates' checks, against this repo
 py tests/run_conformance.py --list # dispositions only, run nothing
 
-py tests/run_e2e.py                # canary only (python-lib, needs API key)
+py tests/run_e2e.py                # canary only (fixture stack, needs API key)
 py tests/run_e2e.py --all          # all tests (live, needs API key)
 py tests/run_e2e.py --area=STK     # run all stack tests
 py tests/run_e2e.py STK-01 FMT-01  # run specific tests by ID
@@ -380,7 +381,7 @@ py tests/run_e2e.py --dry-run      # build prompts, call no model, write no repo
 - CI runs smoke, conformance, `sync.py --check`, `resolve.py --check`,
   the redundancy audit and gitleaks on PRs and on push to main
 - Live e2e calls a model and costs money, so it runs on a cadence
-  rather than per change: the STK-15 canary at every release cut,
+  rather than per change: the STK-24 canary at every release cut,
   recorded in the cut PR, and the full suite at each periodic review
   (PLAYBOOK, "Run the test suite")
 - To validate a new template: run `py tests/run_smoke.py` and

docs/PLAYBOOK.md

@@ -362,7 +362,7 @@ any entry that no longer applies.
 py tests/run_smoke.py              # structural checks — seconds
 py tests/run_conformance.py        # the templates' own checks, run here
 py tests/run_conformance.py --list # dispositions only, run nothing
-py tests/run_e2e.py                # canary test (python-lib)
+py tests/run_e2e.py                # canary test (fixture stack)
 py tests/run_e2e.py --all          # all agent tests
 py tests/run_e2e.py STK-01 FMT-01  # specific tests only
 py tests/run_e2e.py --dry-run      # build prompts, call no model, write no report
@@ -393,7 +393,7 @@ full list of specs. Requires `py -m pip install pyyaml` for the manifest
 check.
 
 `run_e2e.py` is the one runner that calls a model and costs money, so it
-runs on a cadence rather than on every change: the STK-15 canary runs live
+runs on a cadence rather than on every change: the STK-24 canary runs live
 at every release cut, recorded in the cut's pull request, and the full
 suite runs live at each periodic review. It is also the only runner that
 reads what the templates generate — smoke and conformance read the
@@ -831,7 +831,7 @@ below exist to prevent. Run them here anyway.
    pull request's body is this repository's release proposal: it names
    every step of `core/git.md`'s pre-release sequence and the result
    it produced, including the checks that carry no step number here.
-   It also records the STK-15 canary run live against the release commit:
+   It also records the STK-24 canary run live against the release commit:
    the report's tree, provider, model and verdict lines.
    The check's two counts are not required to match: a commit touching
    no template carries no entry, so a journal or tooling change is

docs/SPEC.md

@@ -42,6 +42,16 @@ Extend base templates. Never used directly by a project.
 
 Technology-specific rules. Each stack declares which layers it depends on.
 
+### Fixtures (test input)
+
+`templates/fixture/` holds one stack and one platform that exist for the
+checks, not for projects: `stack-fixture` (layer `fixture`) and
+`platform-fixture`. Each manifest entry carries `fixture: true`, so the
+README, this file's listings, the interview and `generated/` leave them out,
+while the checks that read every stack or platform read them too. They keep
+those checks fed when the concrete stacks and platforms live in forks, and
+the e2e canary (STK-24) runs against the fixture stack.
+
 ### Directory listings
 
 Generated from `templates/manifest.yaml` — run `py tools/sync.py` to refresh.

templates/fixture/platform.md

@@ -0,0 +1,14 @@
+# Platform — Fixture
+[ID: platform-fixture]
+[DEPENDS ON: templates/workflow/quality-gates.md]
+
+Test input for the checks that read a platform, not a platform to pick. It
+resolves as its own root, beside the fixture stack and never inside it.
+
+---
+
+## CI
+[ID: platform-fixture-ci]
+
+- Run every gate the quality-gate tier names on each pull request
+- Fail the pull request when any gate fails; never mark a gate optional

templates/fixture/stack.md

@@ -0,0 +1,35 @@
+# Stack — Fixture
+[DEPENDS ON: templates/security/security.md, templates/workflow/quality-gates.md]
+
+Test input for the checks that read every stack, not a stack to pick. It
+resolves the security and quality-gate tiers and carries each section a
+stack MUST carry, so those checks keep an input in a tree that ships no
+stack of its own.
+
+---
+
+## Stack
+[ID: fixture-stack]
+
+- Language: [any]
+- Service: one HTTP endpoint that returns its request body unchanged
+- Build tool: `make`
+
+---
+
+## Commands
+[ID: fixture-commands]
+
+```
+make install   # install dependencies
+make test      # run the test suite
+make lint      # lint and format check
+```
+
+---
+
+## Project structure
+[ID: fixture-structure]
+
+- `src/` — the service, one module per endpoint
+- `tests/` — one test file per module under `src/`, named after it

templates/manifest.yaml

@@ -10,6 +10,7 @@
 #   label       — human-readable name (stacks only, for interview)
 #   layer       — classification (stacks only, for README table)
 #   produces    — output filename (output entries only)
+#   fixture     — true for test input the checks read and no listing offers
 
 version: "1.0"
 
@@ -142,6 +143,11 @@ platform:
     file: templates/platform/linear.md
     description: Label groups, native priority, sub-issues, code-host sync
     depends_on: [base-issues]
+  - id: platform-fixture
+    file: templates/fixture/platform.md
+    description: Test input for the platform checks; listed nowhere a project reads
+    fixture: true
+    depends_on: [base-quality-gates]
 
 frontend:
   - id: frontend-ux
@@ -224,6 +230,14 @@ backend:
     description: Server-side rendering, partials, escaping, caching, forms, testing
 
 stacks:
+  - id: stack-fixture
+    file: templates/fixture/stack.md
+    label: Fixture
+    layer: fixture
+    description: Test input for the per-stack checks; listed nowhere a project reads
+    fixture: true
+    depends_on: [base-security, base-quality-gates]
+
   - id: stack-htmx
     file: templates/stack/htmx.md
     label: HTMX + server rendering

tests/INDEX.md

@@ -79,6 +79,7 @@ restated in each.
 | `SAIT-E2E-STK-21-001A` | E2E | P1 | Full interview → CLAUDE.md for a NestJS project |
 | `SAIT-E2E-STK-22-001A` | E2E | P1 | Full interview → CLAUDE.md for a C embedded project |
 | `SAIT-E2E-STK-23-001A` | E2E | P1 | Full interview → CLAUDE.md for a tutorial site project |
+| `SAIT-E2E-STK-24-001A` | E2E | P1 | Full interview → CLAUDE.md from the fixture stack (the canary) |
 | `SAIT-E2E-DPL-01-001A` | E2E | P1 | Cloud deployment target produces correct deployment rules |
 | `SAIT-E2E-DPL-02-001A` | E2E | P1 | Hybrid deployment target produces correct deployment rules |
 | `SAIT-E2E-DPL-03-001A` | E2E | P1 | Offline deployment target produces correct deployment rules |

tests/cases.py

@@ -258,9 +258,6 @@
     {
         "id": "STK-15",
         "spec": "SAIT-E2E-STK-15-001A",
-
-        # the canary: the default live test, run at every release cut
-        "canary": True,
         "stack": "templates/stack/python-lib.md",
         "answers": {
             "Project name": "validify",
@@ -408,6 +405,31 @@
         ],
         "forbidden": ["[your project]", "[owner]", "[repo]"],
     },
+    {
+        "id": "STK-24",
+        "spec": "SAIT-E2E-STK-24-001A",
+
+        # the canary: the default live test, run at every release cut. It
+        # reads the fixture stack, which stays in this repository when the
+        # concrete stacks move to forks.
+        "canary": True,
+        "stack": "templates/fixture/stack.md",
+        "answers": {
+            "Project name": "echo",
+            "Language": "Python",
+            "Output format": "CLAUDE.md (inline model)",
+        },
+        "required": [
+            "1. Project", "Commands", "Project structure",
+            "Git", "Testing",
+            "make test",
+            "feat:",
+        ],
+
+        # [any] is the fixture's own placeholder; left in, the agent failed
+        # to substitute the interview answers
+        "forbidden": ["[any]", "[your project]", "[owner]", "[repo]"],
+    },
 ]
 
 # -------------------------------------------------------------------------

tests/context-tiers.txt

@@ -8,6 +8,7 @@
 abstract 200K
 backend 200K
 embedded 128K
+fixture 200K
 hypermedia 128K
 library 200K
 static 200K