feat(site-publish): add split-surface publishing

Authored-By: OpenAI (GPT-5) <noreply@openai.com>

README.md

@@ -92,8 +92,10 @@ routes:
 ```
 
 The caller supplies each declared credential name as an environment variable
-on the action step. Credential values are passed to `aws` only through its
-environment and never appear in a logged command or process argument.
+on the action step. Names must be matched `<NAME>_S3_ACCESS_KEY` and
+`<NAME>_S3_SECRET_KEY` pairs; arbitrary environment variables cannot become
+publication credentials. Values pass to `aws` only through its environment and
+never appear in a logged command or process argument.
 
 ```yaml
       - uses: https://code.fritzlab.net/action/site-publish@v1
@@ -127,6 +129,12 @@ writes identical even though Garage v2.2.0 has no conditional destination
 write. An identical retry converges; a changed object, missing digest metadata,
 wrong address, or nested policy under that immutable prefix fails publication.
 
+Artifact input directories must be pairwise disjoint after filesystem
+resolution. Publication stops before build or upload if one contains another or
+escapes the repository, preventing protected input from entering a public
+artifact. Split storage endpoints are pinned to Garage, and each website
+authority is derived from its bucket; a site cannot expose an arbitrary backend.
+
 Each split route gets a bucket-specific `<bucket>.web.sjc001.fritzlab.net`
 ExternalName Service annotated to disable pass-host-header and a separate Ingress. Route
 Ingresses share the hostname's certificate Secret. The access middleware and

scripts/build.py

@@ -5,7 +5,7 @@ import subprocess
 import tempfile
 from pathlib import Path
 
-from utils import EXCLUDE_FILES, env, parse_site_yaml, run
+from utils import EXCLUDE_FILES, env, parse_site_yaml, run, validate_artifact_inputs
 
 
 def build_artifact(site_dir, artifact):
@@ -64,5 +64,7 @@ def cmd_build():
         print("Site disabled — skipping build")
         return
 
+    validate_artifact_inputs(site_dir, cfg)
+
     for artifact in cfg["artifacts"]:
         build_artifact(site_dir, artifact)

scripts/deploy.py

@@ -23,6 +23,7 @@ from utils import (
     parse_site_yaml,
     render_templates,
     run,
+    validate_artifact_inputs,
 )
 
 GARAGE_ADMIN_ENDPOINT = os.environ.get(
@@ -350,4 +351,5 @@ def cmd_deploy():
         decommission(site_name, token, [artifact["bucket"] for artifact in cfg["artifacts"]])
         return
 
+    validate_artifact_inputs(site_dir, cfg)
     deploy_static(site_name, site_dir, action_dir, token, cfg)

scripts/utils.py

@@ -26,6 +26,7 @@ EXCLUDE_FILES = {
 VALID_TYPES = {"static", "hugo", "mkdocs"}
 NAME_RE = re.compile(r"^[a-z0-9](?:[a-z0-9-]{0,61}[a-z0-9])?$")
 ENV_RE = re.compile(r"^[A-Z_][A-Z0-9_]*$")
+ACCESS_KEY_ENV_RE = re.compile(r"^([A-Z][A-Z0-9_]*)_S3_ACCESS_KEY$")
 BUCKET_RE = re.compile(r"^[a-z0-9](?:[a-z0-9.-]{1,61}[a-z0-9])?$")
 MIDDLEWARE_RE = re.compile(r"^[A-Za-z0-9](?:[A-Za-z0-9_-]{0,61}[A-Za-z0-9])?$")
 
@@ -238,7 +239,7 @@ def _artifact(item, index):
     if not isinstance(name, str) or not NAME_RE.fullmatch(name):
         raise ConfigError(f"{label}.name must be a DNS label")
     publish = _mapping(item.get("publish"), f"{label}.publish")
-    _known_keys(publish, {"bucket", "endpoint", "website_authority", "credentials"}, f"{label}.publish")
+    _known_keys(publish, {"bucket", "credentials"}, f"{label}.publish")
     bucket = publish.get("bucket")
     if not isinstance(bucket, str) or not BUCKET_RE.fullmatch(bucket):
         raise ConfigError(f"{label}.publish.bucket is not a valid bucket name")
@@ -250,6 +251,15 @@ def _artifact(item, index):
         if not isinstance(value, str) or not ENV_RE.fullmatch(value):
             raise ConfigError(f"{label}.publish.credentials.{key} must name an environment variable")
         normalized_credentials[key] = value
+    access_match = ACCESS_KEY_ENV_RE.fullmatch(normalized_credentials["access_key_env"])
+    expected_secret = (
+        f"{access_match.group(1)}_S3_SECRET_KEY" if access_match else None
+    )
+    if normalized_credentials["secret_key_env"] != expected_secret:
+        raise ConfigError(
+            f"{label}.publish.credentials must be a matched "
+            "<NAME>_S3_ACCESS_KEY and <NAME>_S3_SECRET_KEY pair"
+        )
     cache = _mapping(item.get("cache"), f"{label}.cache")
     _known_keys(cache, {"rules"}, f"{label}.cache")
     rules = _list(cache.get("rules"), f"{label}.cache.rules")
@@ -278,10 +288,7 @@ def _artifact(item, index):
     for path in immutable_paths:
         if any(other.startswith(f"{path}/") for other in paths):
             raise ConfigError(f"{label}.cache.rules cannot nest another policy under immutable /{path}")
-    authority = _hostname(
-        publish.get("website_authority", f"{bucket}.{DEFAULT_WEBSITE_SUFFIX}"),
-        f"{label}.publish.website_authority",
-    )
+    authority = f"{bucket}.{DEFAULT_WEBSITE_SUFFIX}"
     if not isinstance(item.get("tidy", True), bool):
         raise ConfigError(f"{label}.tidy must be a boolean")
     return {
@@ -292,7 +299,7 @@ def _artifact(item, index):
         "excludes": _strings(item.get("excludes") or [], f"{label}.excludes"),
         "build_dir": f".site-publish/{name}/html",
         "bucket": bucket,
-        "s3_endpoint": _endpoint(publish.get("endpoint", DEFAULT_S3_ENDPOINT), f"{label}.publish.endpoint"),
+        "s3_endpoint": DEFAULT_S3_ENDPOINT,
         "website_authority": authority,
         "credentials": normalized_credentials,
         "cache_rules": cache_rules,
@@ -418,6 +425,25 @@ def normalize_site_config(raw, site_name):
     return cfg
 
 
+def validate_artifact_inputs(site_dir, cfg):
+    """Reject source containment before a public or protected build starts."""
+    root = Path(site_dir).resolve()
+    sources = []
+    for artifact in cfg["artifacts"]:
+        source = (root / artifact["content_dir"]).resolve()
+        if source != root and root not in source.parents:
+            raise ConfigError(
+                f"artifact {artifact['name']} content_dir resolves outside the repository"
+            )
+        sources.append((artifact["name"], source))
+    for index, (name, source) in enumerate(sources):
+        for other_name, other_source in sources[index + 1:]:
+            if source == other_source or source in other_source.parents or other_source in source.parents:
+                raise ConfigError(
+                    f"artifact build inputs overlap after resolution: {name} and {other_name}"
+                )
+
+
 def parse_site_yaml(site_dir, site_name=None):
     path = Path(site_dir) / "site.yaml"
     if not path.exists():

tests/test_contract.py

@@ -24,7 +24,7 @@ sys.path.insert(0, str(ROOT / "scripts"))
 import deploy
 import build
 import utils
-from utils import ConfigError, normalize_site_config
+from utils import ConfigError, normalize_site_config, validate_artifact_inputs
 
 
 def fixture(name):
@@ -151,12 +151,11 @@ class ConfigContractTests(unittest.TestCase):
         )
 
     def test_publication_credentials_cannot_be_reused(self):
-        self.assert_invalid(
-            lambda raw: raw["artifacts"][0]["publish"]["credentials"].__setitem__(
-                "access_key_env", "DIST_S3_ACCESS_KEY"
-            ),
-            "publication credential DIST_S3_ACCESS_KEY is reused",
-        )
+        def mutate(raw):
+            raw["artifacts"][0]["publish"]["credentials"] = copy.deepcopy(
+                raw["artifacts"][1]["publish"]["credentials"]
+            )
+        self.assert_invalid(mutate, "publication credential DIST_S3_ACCESS_KEY is reused")
 
     def test_cache_directive_contradiction_is_rejected(self):
         self.assert_invalid(
@@ -191,6 +190,35 @@ class ConfigContractTests(unittest.TestCase):
             "unknown fields: storage_bucket",
         )
 
+    def test_backend_authority_and_endpoint_are_derived(self):
+        for field, value in (
+            ("endpoint", "https://attacker.example"),
+            ("website_authority", "internal-api.default.svc.k8s.sjc001.fritzlab.net"),
+        ):
+            with self.subTest(field=field):
+                self.assert_invalid(
+                    lambda raw, field=field, value=value: raw["artifacts"][0]["publish"].__setitem__(field, value),
+                    f"unknown fields: {field}",
+                )
+
+    def test_publication_credentials_use_dedicated_matched_names(self):
+        self.assert_invalid(
+            lambda raw: raw["artifacts"][0]["publish"]["credentials"].__setitem__(
+                "access_key_env", "CI_BOT_TOKEN"
+            ),
+            "must be a matched <NAME>_S3_ACCESS_KEY",
+        )
+
+    def test_resolved_build_inputs_must_be_pairwise_disjoint(self):
+        raw = copy.deepcopy(self.raw)
+        raw["artifacts"][1]["content_dir"] = ""
+        cfg = normalize_site_config(raw, "baseline.fritzlab.net")
+        with tempfile.TemporaryDirectory() as tmp:
+            root = Path(tmp)
+            (root / "portal" / "build").mkdir(parents=True)
+            with self.assertRaisesRegex(ConfigError, "build inputs overlap after resolution"):
+                validate_artifact_inputs(root, cfg)
+
 
 class GenerationTests(unittest.TestCase):
     def render(self, raw):