From af89c3aa646c198cce4cd2b11fb98e076a757578 Mon Sep 17 00:00:00 2001
From: nvk <git@nvk.org>
Date: Mon, 24 Aug 2026 19:09:14 -0400
Subject: [PATCH] Fix raw source path linting without provenance false
 positives

---
 AGENTS.md                                     |  4 +-
 claude-plugin/bin/llm-wiki                    | 77 +++++++++++++++++++
 .../skills/wiki-manager/references/linting.md |  8 ++
 .../wiki-manager/references/wiki-structure.md |  2 +-
 plugins/llm-wiki-opencode/bin/llm-wiki        | 77 +++++++++++++++++++
 plugins/llm-wiki/bin/llm-wiki                 | 77 +++++++++++++++++++
 .../skills/wiki/references/linting.md         |  8 ++
 .../skills/wiki/references/wiki-structure.md  |  2 +-
 scripts/llm-wiki                              | 77 +++++++++++++++++++
 .../defects/raw-source-unresolved/_index.md   | 36 +++++++++
 .../defects/raw-source-unresolved/config.md   | 15 ++++
 .../raw-source-unresolved/datasets/_index.md  | 18 +++++
 .../bitcointalk-temporal-graph/MANIFEST.md    | 48 ++++++++++++
 .../bitcointalk-temporal-graph/_index.md      | 17 ++++
 .../profiles/_index.md                        | 10 +++
 .../queries/_index.md                         | 10 +++
 .../samples/_index.md                         | 10 +++
 .../raw-source-unresolved/inventory/_index.md | 30 ++++++++
 .../inventory/candidates/_index.md            | 11 +++
 .../candidates/bitcointalk-archive.md         | 33 ++++++++
 .../inventory/corpora/_index.md               | 10 +++
 .../inventory/entities/_index.md              | 10 +++
 .../inventory/items/_index.md                 | 11 +++
 .../inventory/items/trx4m-ring-and-pinion.md  | 44 +++++++++++
 .../inventory/views/_index.md                 | 11 +++
 .../inventory/views/active-actions.md         | 18 +++++
 .../defects/raw-source-unresolved/log.md      | 10 +++
 .../raw-source-unresolved/output/_index.md    |  5 ++
 .../output/sample-output.md                   | 12 +++
 .../raw-source-unresolved/raw/_index.md       | 20 +++++
 .../raw/articles/2026-01-01-sample-article.md | 14 ++++
 .../raw/articles/2026-01-02-second-article.md | 14 ++++
 .../articles/2026-01-03-Title Cased Source.md | 14 ++++
 .../raw/articles/_index.md                    |  7 ++
 .../raw-source-unresolved/raw/data/_index.md  |  4 +
 .../raw-source-unresolved/raw/notes/_index.md |  4 +
 .../raw/papers/2026-01-01-sample-paper.md     | 14 ++++
 .../raw/papers/_index.md                      |  5 ++
 .../raw-source-unresolved/raw/repos/_index.md |  4 +
 .../defects/raw-source-unresolved/schema.md   | 55 +++++++++++++
 .../raw-source-unresolved/wiki/_index.md      | 13 ++++
 .../wiki/concepts/_index.md                   |  5 ++
 .../wiki/concepts/sample-concept.md           | 31 ++++++++
 .../wiki/references/_index.md                 |  5 ++
 .../wiki/references/sample-reference.md       | 31 ++++++++
 .../wiki/theses/_index.md                     |  4 +
 .../wiki/topics/_index.md                     |  4 +
 tests/generate-defect-fixtures.sh             |  7 ++
 tests/test-local-cli-lint.sh                  | 28 +++++++
 tests/test-structure.sh                       |  6 ++
 50 files changed, 986 insertions(+), 4 deletions(-)
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/config.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/datasets/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/MANIFEST.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/profiles/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/queries/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/samples/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/candidates/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/candidates/bitcointalk-archive.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/corpora/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/entities/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/items/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/items/trx4m-ring-and-pinion.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/views/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/inventory/views/active-actions.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/log.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/output/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/output/sample-output.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-01-sample-article.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-02-second-article.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-03-Title Cased Source.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/articles/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/data/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/notes/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/papers/2026-01-01-sample-paper.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/papers/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/raw/repos/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/schema.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/wiki/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/wiki/concepts/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/wiki/concepts/sample-concept.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/wiki/references/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/wiki/references/sample-reference.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/wiki/theses/_index.md
 create mode 100644 tests/fixtures/defects/raw-source-unresolved/wiki/topics/_index.md

diff --git a/AGENTS.md b/AGENTS.md
index 84de5d4..1dccaa3 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -211,7 +211,7 @@ established wikis and copy only the useful proposal parts into `schema.md`.
 ```yaml
 ---
 title: "Title"
-source: "URL or filepath or MANUAL"
+source: "URL, filepath, MANUAL, or provenance label such as session"
 type: articles|papers|repos|notes|data
 ingested: YYYY-MM-DD
 tags: [tag1, tag2]
@@ -854,7 +854,7 @@ Health checks with auto-fix capability. Lint **is** the migration path — there
   path/status drift. Normal lint reports archived topics as skipped;
   `--include-archived` or `--archived-only` structurally maintains them.
 
-**Checks**: structure integrity, frontmatter validity (plus legacy key/value aliases C13), canonical placement of raw/wiki files (C11), unknown-file quarantine for raw/wiki/inventory/datasets/root (C12), index consistency, link integrity, source provenance (dangling refs, unresolved retraction markers), tag hygiene, coverage, project `WHY.md` presence (C8a), project staleness via source chain (C8b), legacy `_project.md` migration to `WHY.md` (C8c), project candidates (C9), inventory migration candidates (C16), dataset migration candidates (C17), archive registry drift and active/archive collisions (C19), deep fact-checking (optional).
+**Checks**: structure integrity, frontmatter validity (plus legacy key/value aliases C13), canonical placement of raw/wiki files (C11), unknown-file quarantine for raw/wiki/inventory/datasets/root (C12), index consistency, link integrity, source provenance (dangling refs, unresolved explicit local raw-source paths, unresolved retraction markers), tag hygiene, coverage, project `WHY.md` presence (C8a), project staleness via source chain (C8b), legacy `_project.md` migration to `WHY.md` (C8c), project candidates (C9), inventory migration candidates (C16), dataset migration candidates (C17), archive registry drift and active/archive collisions (C19), deep fact-checking (optional).
 
 **Auto-fix** (`--fix`): rewrite legacy frontmatter keys/values to canonical (C13), move misplaced raw/wiki files to their canonical directory (C11), quarantine unknown files to `inbox/.unknown` (C12), migrate legacy `_project.md` to `WHY.md` (C8c), add a default human-owned `schema.md` in advisory mode when missing, repair missing indexes inside existing inventory/dataset layers (C16/C17), repair unambiguous archive registry path/status drift (C19), missing indexes, orphan files, dead index entries, statistics mismatch, missing bidirectional links, empty frontmatter fields, dangling source references, regenerate projects-aware `output/_index.md`. Never auto-delete unknown directories. Never auto-create `WHY.md` with placeholder goals (C8a is warn-only — manufactured rationale is worse than missing). Never create completely absent optional inventory or dataset trees just to populate placeholders. Never auto-move files into projects (C9 is human-authored via `/wiki:project`). Never auto-migrate output artifacts into inventory or dataset records (C16/C17 are explicit via `/wiki:inventory migrate-output --apply` and `/wiki:dataset migrate-output --apply`). Never move topics into or out of archive during lint; archive/restore is explicit. On slug collisions during a placement move, skip and warn.
 
diff --git a/claude-plugin/bin/llm-wiki b/claude-plugin/bin/llm-wiki
index 0d8a936..304c2a0 100755
--- a/claude-plugin/bin/llm-wiki
+++ b/claude-plugin/bin/llm-wiki
@@ -60,6 +60,9 @@ CONFIDENCE_VALUES = {"high", "medium", "low"}
 VOLATILITY_VALUES = {"hot", "warm", "cold"}
 PERMISSION_DENIED_ERRNOS = {1, 13}
 WINDOWS_ABSOLUTE_PATH_RE = re.compile(r"^/?[A-Za-z]:[\\/]")
+SCHEMELESS_WEB_SOURCE_RE = re.compile(
+    r"^(?:[A-Za-z0-9-]+\.)+[A-Za-z]{2,}(?::\d+)?(?:[/#?]|$)"
+)
 ADAPTER_PROTOCOL = "llm-wiki-adapter/v1"
 ADAPTER_REGISTRY_SCHEMA = 1
 ADAPTER_MANIFEST_NAME = ".llm-wiki-adapter.json"
@@ -1473,6 +1476,16 @@ def check_source_provenance(ctx: LintContext) -> None:
                             f"Inventory source reference does not resolve: {source}.",
                             doc.path,
                         )
+        elif rel.parts[0] == "raw":
+            source = doc.frontmatter.get("source")
+            if source:
+                candidate = raw_source_path_candidate(ctx, doc.path, str(source))
+                if candidate is not None and not source_path_exists(candidate):
+                    ctx.issue(
+                        "warning",
+                        f"Raw source reference does not resolve: {source}.",
+                        doc.path,
+                    )
 
         if "RETRACTED-SOURCE" in doc.body:
             ctx.issue("warning", "Retracted-source marker remains in the file body.", doc.path)
@@ -1562,6 +1575,70 @@ def resolve_source_ref(ctx: LintContext, owner: Path, ref: str, wiki_source: boo
     return None
 
 
+def raw_source_path_candidate(ctx: LintContext, owner: Path, ref: str) -> Path | None:
+    """Return a candidate only when a raw source value is clearly a local path.
+
+    Raw provenance also uses sentinels such as ``session`` and descriptive
+    labels. Treating every non-HTTP value as a path would make those established
+    values fail lint. URI schemes remain external except for ``file://``; bare
+    relative paths must contain a separator and no whitespace to be considered
+    unambiguous.
+    """
+    ref = strip_matching_quotes(ref.strip())
+    if not ref:
+        return None
+
+    windows_ref = windows_absolute_path(ref)
+    if windows_ref is not None:
+        return windows_ref
+
+    try:
+        parsed = urllib.parse.urlsplit(ref)
+    except ValueError:
+        parsed = None
+    if parsed is not None and parsed.scheme:
+        # Do not mistake a Windows drive letter for a URI scheme on a platform
+        # where Windows paths are not native.
+        if WINDOWS_ABSOLUTE_PATH_RE.match(ref):
+            return Path(ref)
+        if parsed.scheme.lower() != "file":
+            return None
+        path_text = urllib.parse.unquote(parsed.path)
+        if parsed.netloc and parsed.netloc.lower() != "localhost":
+            path_text = f"//{parsed.netloc}{path_text}"
+        if os.name == "nt" and re.match(r"^/[A-Za-z]:/", path_text):
+            path_text = path_text[1:]
+        file_path = Path(path_text)
+        return file_path if file_path.is_absolute() else ctx.root / file_path
+
+    if SCHEMELESS_WEB_SOURCE_RE.match(ref):
+        return None
+
+    explicit_relative = ref.startswith(("../", "./", "..\\", ".\\"))
+    home_relative = ref == "~" or ref.startswith(("~/", "~\\"))
+    absolute = Path(ref).is_absolute()
+    unambiguous_bare_relative = (
+        not any(char.isspace() for char in ref) and ("/" in ref or "\\" in ref)
+    )
+    if not (explicit_relative or home_relative or absolute or unambiguous_bare_relative):
+        return None
+
+    ref_path = expand_leading_tilde(ref)
+    if ref_path.is_absolute():
+        return ref_path
+    if explicit_relative:
+        return owner.parent / ref_path
+    return ctx.root / ref_path
+
+
+def source_path_exists(path: Path) -> bool:
+    """Check a source file or directory without crashing on sandbox denials."""
+    try:
+        return path.exists()
+    except OSError:
+        return False
+
+
 def strip_matching_quotes(value: str) -> str:
     if (value.startswith('"') and value.endswith('"')) or (
         value.startswith("'") and value.endswith("'")
diff --git a/claude-plugin/skills/wiki-manager/references/linting.md b/claude-plugin/skills/wiki-manager/references/linting.md
index bcc3f04..20311b1 100644
--- a/claude-plugin/skills/wiki-manager/references/linting.md
+++ b/claude-plugin/skills/wiki-manager/references/linting.md
@@ -102,6 +102,8 @@ There is no `/wiki:migrate` command and there should never be one. Lint rules **
 
 - [ ] All markdown links `[text](path)` in wiki articles and inventory records
   resolve to existing local files when they are local paths
+- [ ] Do not apply this check to raw source bodies. Imported Markdown may retain
+  relative links whose targets live only in the upstream source collection.
 - [ ] All "See Also" links are bidirectional (if A→B, then B→A)
 - [ ] All "Sources" links in wiki articles point to existing raw files. Links to paths with spaces should use angle-bracket markdown destinations, e.g. `[Title](<../../raw/articles/File Name.md>)`.
 
@@ -112,6 +114,12 @@ There is no `/wiki:migrate` command and there should never be one. Lint rules **
   existing files under `raw/`, `wiki/`, `output/`, `datasets/`, or `inventory/`.
   External URLs are allowed. Inventory provenance is operational state and must
   not be treated as factual evidence for compile/query/audit verdicts.
+- [ ] Explicit local paths in a raw source's scalar `source:` field resolve to
+  an existing file or directory. Treat absolute paths, `file://` URIs,
+  `./`/`../`/`~/` paths, and whitespace-free relative paths containing a
+  directory separator as local. Other URI schemes, schemeless web URLs,
+  sentinels such as `MANUAL` or `session`, and free-form provenance labels are
+  not local-path checks.
 - [ ] No `<!--RETRACTED-SOURCE-->` markers remain in article body (these should be resolved via `--recompile` or manual review)
 - [ ] No raw source file is referenced by zero wiki articles (orphan source — suggest compilation or removal)
 - [ ] Exempt raw files tagged `collection-manifest` from orphan-source warnings. A collection manifest is operational provenance for a batch import; child sources should be compiled, but the manifest itself does not need to appear in article `sources:`.
diff --git a/claude-plugin/skills/wiki-manager/references/wiki-structure.md b/claude-plugin/skills/wiki-manager/references/wiki-structure.md
index 8f4f1bb..edbde7a 100644
--- a/claude-plugin/skills/wiki-manager/references/wiki-structure.md
+++ b/claude-plugin/skills/wiki-manager/references/wiki-structure.md
@@ -372,7 +372,7 @@ Topic-guide helpers:
 ```markdown
 ---
 title: "Title"
-source: "URL or filepath or MANUAL"
+source: "URL, filepath, MANUAL, or provenance label such as session"
 type: articles|papers|repos|notes|data
 ingested: YYYY-MM-DD
 tags: [tag1, tag2]
diff --git a/plugins/llm-wiki-opencode/bin/llm-wiki b/plugins/llm-wiki-opencode/bin/llm-wiki
index 0d8a936..304c2a0 100755
--- a/plugins/llm-wiki-opencode/bin/llm-wiki
+++ b/plugins/llm-wiki-opencode/bin/llm-wiki
@@ -60,6 +60,9 @@ CONFIDENCE_VALUES = {"high", "medium", "low"}
 VOLATILITY_VALUES = {"hot", "warm", "cold"}
 PERMISSION_DENIED_ERRNOS = {1, 13}
 WINDOWS_ABSOLUTE_PATH_RE = re.compile(r"^/?[A-Za-z]:[\\/]")
+SCHEMELESS_WEB_SOURCE_RE = re.compile(
+    r"^(?:[A-Za-z0-9-]+\.)+[A-Za-z]{2,}(?::\d+)?(?:[/#?]|$)"
+)
 ADAPTER_PROTOCOL = "llm-wiki-adapter/v1"
 ADAPTER_REGISTRY_SCHEMA = 1
 ADAPTER_MANIFEST_NAME = ".llm-wiki-adapter.json"
@@ -1473,6 +1476,16 @@ def check_source_provenance(ctx: LintContext) -> None:
                             f"Inventory source reference does not resolve: {source}.",
                             doc.path,
                         )
+        elif rel.parts[0] == "raw":
+            source = doc.frontmatter.get("source")
+            if source:
+                candidate = raw_source_path_candidate(ctx, doc.path, str(source))
+                if candidate is not None and not source_path_exists(candidate):
+                    ctx.issue(
+                        "warning",
+                        f"Raw source reference does not resolve: {source}.",
+                        doc.path,
+                    )
 
         if "RETRACTED-SOURCE" in doc.body:
             ctx.issue("warning", "Retracted-source marker remains in the file body.", doc.path)
@@ -1562,6 +1575,70 @@ def resolve_source_ref(ctx: LintContext, owner: Path, ref: str, wiki_source: boo
     return None
 
 
+def raw_source_path_candidate(ctx: LintContext, owner: Path, ref: str) -> Path | None:
+    """Return a candidate only when a raw source value is clearly a local path.
+
+    Raw provenance also uses sentinels such as ``session`` and descriptive
+    labels. Treating every non-HTTP value as a path would make those established
+    values fail lint. URI schemes remain external except for ``file://``; bare
+    relative paths must contain a separator and no whitespace to be considered
+    unambiguous.
+    """
+    ref = strip_matching_quotes(ref.strip())
+    if not ref:
+        return None
+
+    windows_ref = windows_absolute_path(ref)
+    if windows_ref is not None:
+        return windows_ref
+
+    try:
+        parsed = urllib.parse.urlsplit(ref)
+    except ValueError:
+        parsed = None
+    if parsed is not None and parsed.scheme:
+        # Do not mistake a Windows drive letter for a URI scheme on a platform
+        # where Windows paths are not native.
+        if WINDOWS_ABSOLUTE_PATH_RE.match(ref):
+            return Path(ref)
+        if parsed.scheme.lower() != "file":
+            return None
+        path_text = urllib.parse.unquote(parsed.path)
+        if parsed.netloc and parsed.netloc.lower() != "localhost":
+            path_text = f"//{parsed.netloc}{path_text}"
+        if os.name == "nt" and re.match(r"^/[A-Za-z]:/", path_text):
+            path_text = path_text[1:]
+        file_path = Path(path_text)
+        return file_path if file_path.is_absolute() else ctx.root / file_path
+
+    if SCHEMELESS_WEB_SOURCE_RE.match(ref):
+        return None
+
+    explicit_relative = ref.startswith(("../", "./", "..\\", ".\\"))
+    home_relative = ref == "~" or ref.startswith(("~/", "~\\"))
+    absolute = Path(ref).is_absolute()
+    unambiguous_bare_relative = (
+        not any(char.isspace() for char in ref) and ("/" in ref or "\\" in ref)
+    )
+    if not (explicit_relative or home_relative or absolute or unambiguous_bare_relative):
+        return None
+
+    ref_path = expand_leading_tilde(ref)
+    if ref_path.is_absolute():
+        return ref_path
+    if explicit_relative:
+        return owner.parent / ref_path
+    return ctx.root / ref_path
+
+
+def source_path_exists(path: Path) -> bool:
+    """Check a source file or directory without crashing on sandbox denials."""
+    try:
+        return path.exists()
+    except OSError:
+        return False
+
+
 def strip_matching_quotes(value: str) -> str:
     if (value.startswith('"') and value.endswith('"')) or (
         value.startswith("'") and value.endswith("'")
diff --git a/plugins/llm-wiki/bin/llm-wiki b/plugins/llm-wiki/bin/llm-wiki
index 0d8a936..304c2a0 100755
--- a/plugins/llm-wiki/bin/llm-wiki
+++ b/plugins/llm-wiki/bin/llm-wiki
@@ -60,6 +60,9 @@ CONFIDENCE_VALUES = {"high", "medium", "low"}
 VOLATILITY_VALUES = {"hot", "warm", "cold"}
 PERMISSION_DENIED_ERRNOS = {1, 13}
 WINDOWS_ABSOLUTE_PATH_RE = re.compile(r"^/?[A-Za-z]:[\\/]")
+SCHEMELESS_WEB_SOURCE_RE = re.compile(
+    r"^(?:[A-Za-z0-9-]+\.)+[A-Za-z]{2,}(?::\d+)?(?:[/#?]|$)"
+)
 ADAPTER_PROTOCOL = "llm-wiki-adapter/v1"
 ADAPTER_REGISTRY_SCHEMA = 1
 ADAPTER_MANIFEST_NAME = ".llm-wiki-adapter.json"
@@ -1473,6 +1476,16 @@ def check_source_provenance(ctx: LintContext) -> None:
                             f"Inventory source reference does not resolve: {source}.",
                             doc.path,
                         )
+        elif rel.parts[0] == "raw":
+            source = doc.frontmatter.get("source")
+            if source:
+                candidate = raw_source_path_candidate(ctx, doc.path, str(source))
+                if candidate is not None and not source_path_exists(candidate):
+                    ctx.issue(
+                        "warning",
+                        f"Raw source reference does not resolve: {source}.",
+                        doc.path,
+                    )
 
         if "RETRACTED-SOURCE" in doc.body:
             ctx.issue("warning", "Retracted-source marker remains in the file body.", doc.path)
@@ -1562,6 +1575,70 @@ def resolve_source_ref(ctx: LintContext, owner: Path, ref: str, wiki_source: boo
     return None
 
 
+def raw_source_path_candidate(ctx: LintContext, owner: Path, ref: str) -> Path | None:
+    """Return a candidate only when a raw source value is clearly a local path.
+
+    Raw provenance also uses sentinels such as ``session`` and descriptive
+    labels. Treating every non-HTTP value as a path would make those established
+    values fail lint. URI schemes remain external except for ``file://``; bare
+    relative paths must contain a separator and no whitespace to be considered
+    unambiguous.
+    """
+    ref = strip_matching_quotes(ref.strip())
+    if not ref:
+        return None
+
+    windows_ref = windows_absolute_path(ref)
+    if windows_ref is not None:
+        return windows_ref
+
+    try:
+        parsed = urllib.parse.urlsplit(ref)
+    except ValueError:
+        parsed = None
+    if parsed is not None and parsed.scheme:
+        # Do not mistake a Windows drive letter for a URI scheme on a platform
+        # where Windows paths are not native.
+        if WINDOWS_ABSOLUTE_PATH_RE.match(ref):
+            return Path(ref)
+        if parsed.scheme.lower() != "file":
+            return None
+        path_text = urllib.parse.unquote(parsed.path)
+        if parsed.netloc and parsed.netloc.lower() != "localhost":
+            path_text = f"//{parsed.netloc}{path_text}"
+        if os.name == "nt" and re.match(r"^/[A-Za-z]:/", path_text):
+            path_text = path_text[1:]
+        file_path = Path(path_text)
+        return file_path if file_path.is_absolute() else ctx.root / file_path
+
+    if SCHEMELESS_WEB_SOURCE_RE.match(ref):
+        return None
+
+    explicit_relative = ref.startswith(("../", "./", "..\\", ".\\"))
+    home_relative = ref == "~" or ref.startswith(("~/", "~\\"))
+    absolute = Path(ref).is_absolute()
+    unambiguous_bare_relative = (
+        not any(char.isspace() for char in ref) and ("/" in ref or "\\" in ref)
+    )
+    if not (explicit_relative or home_relative or absolute or unambiguous_bare_relative):
+        return None
+
+    ref_path = expand_leading_tilde(ref)
+    if ref_path.is_absolute():
+        return ref_path
+    if explicit_relative:
+        return owner.parent / ref_path
+    return ctx.root / ref_path
+
+
+def source_path_exists(path: Path) -> bool:
+    """Check a source file or directory without crashing on sandbox denials."""
+    try:
+        return path.exists()
+    except OSError:
+        return False
+
+
 def strip_matching_quotes(value: str) -> str:
     if (value.startswith('"') and value.endswith('"')) or (
         value.startswith("'") and value.endswith("'")
diff --git a/plugins/llm-wiki/skills/wiki/references/linting.md b/plugins/llm-wiki/skills/wiki/references/linting.md
index bcc3f04..20311b1 100644
--- a/plugins/llm-wiki/skills/wiki/references/linting.md
+++ b/plugins/llm-wiki/skills/wiki/references/linting.md
@@ -102,6 +102,8 @@ There is no `/wiki:migrate` command and there should never be one. Lint rules **
 
 - [ ] All markdown links `[text](path)` in wiki articles and inventory records
   resolve to existing local files when they are local paths
+- [ ] Do not apply this check to raw source bodies. Imported Markdown may retain
+  relative links whose targets live only in the upstream source collection.
 - [ ] All "See Also" links are bidirectional (if A→B, then B→A)
 - [ ] All "Sources" links in wiki articles point to existing raw files. Links to paths with spaces should use angle-bracket markdown destinations, e.g. `[Title](<../../raw/articles/File Name.md>)`.
 
@@ -112,6 +114,12 @@ There is no `/wiki:migrate` command and there should never be one. Lint rules **
   existing files under `raw/`, `wiki/`, `output/`, `datasets/`, or `inventory/`.
   External URLs are allowed. Inventory provenance is operational state and must
   not be treated as factual evidence for compile/query/audit verdicts.
+- [ ] Explicit local paths in a raw source's scalar `source:` field resolve to
+  an existing file or directory. Treat absolute paths, `file://` URIs,
+  `./`/`../`/`~/` paths, and whitespace-free relative paths containing a
+  directory separator as local. Other URI schemes, schemeless web URLs,
+  sentinels such as `MANUAL` or `session`, and free-form provenance labels are
+  not local-path checks.
 - [ ] No `<!--RETRACTED-SOURCE-->` markers remain in article body (these should be resolved via `--recompile` or manual review)
 - [ ] No raw source file is referenced by zero wiki articles (orphan source — suggest compilation or removal)
 - [ ] Exempt raw files tagged `collection-manifest` from orphan-source warnings. A collection manifest is operational provenance for a batch import; child sources should be compiled, but the manifest itself does not need to appear in article `sources:`.
diff --git a/plugins/llm-wiki/skills/wiki/references/wiki-structure.md b/plugins/llm-wiki/skills/wiki/references/wiki-structure.md
index 8f4f1bb..edbde7a 100644
--- a/plugins/llm-wiki/skills/wiki/references/wiki-structure.md
+++ b/plugins/llm-wiki/skills/wiki/references/wiki-structure.md
@@ -372,7 +372,7 @@ Topic-guide helpers:
 ```markdown
 ---
 title: "Title"
-source: "URL or filepath or MANUAL"
+source: "URL, filepath, MANUAL, or provenance label such as session"
 type: articles|papers|repos|notes|data
 ingested: YYYY-MM-DD
 tags: [tag1, tag2]
diff --git a/scripts/llm-wiki b/scripts/llm-wiki
index 0d8a936..304c2a0 100755
--- a/scripts/llm-wiki
+++ b/scripts/llm-wiki
@@ -60,6 +60,9 @@ CONFIDENCE_VALUES = {"high", "medium", "low"}
 VOLATILITY_VALUES = {"hot", "warm", "cold"}
 PERMISSION_DENIED_ERRNOS = {1, 13}
 WINDOWS_ABSOLUTE_PATH_RE = re.compile(r"^/?[A-Za-z]:[\\/]")
+SCHEMELESS_WEB_SOURCE_RE = re.compile(
+    r"^(?:[A-Za-z0-9-]+\.)+[A-Za-z]{2,}(?::\d+)?(?:[/#?]|$)"
+)
 ADAPTER_PROTOCOL = "llm-wiki-adapter/v1"
 ADAPTER_REGISTRY_SCHEMA = 1
 ADAPTER_MANIFEST_NAME = ".llm-wiki-adapter.json"
@@ -1473,6 +1476,16 @@ def check_source_provenance(ctx: LintContext) -> None:
                             f"Inventory source reference does not resolve: {source}.",
                             doc.path,
                         )
+        elif rel.parts[0] == "raw":
+            source = doc.frontmatter.get("source")
+            if source:
+                candidate = raw_source_path_candidate(ctx, doc.path, str(source))
+                if candidate is not None and not source_path_exists(candidate):
+                    ctx.issue(
+                        "warning",
+                        f"Raw source reference does not resolve: {source}.",
+                        doc.path,
+                    )
 
         if "RETRACTED-SOURCE" in doc.body:
             ctx.issue("warning", "Retracted-source marker remains in the file body.", doc.path)
@@ -1562,6 +1575,70 @@ def resolve_source_ref(ctx: LintContext, owner: Path, ref: str, wiki_source: boo
     return None
 
 
+def raw_source_path_candidate(ctx: LintContext, owner: Path, ref: str) -> Path | None:
+    """Return a candidate only when a raw source value is clearly a local path.
+
+    Raw provenance also uses sentinels such as ``session`` and descriptive
+    labels. Treating every non-HTTP value as a path would make those established
+    values fail lint. URI schemes remain external except for ``file://``; bare
+    relative paths must contain a separator and no whitespace to be considered
+    unambiguous.
+    """
+    ref = strip_matching_quotes(ref.strip())
+    if not ref:
+        return None
+
+    windows_ref = windows_absolute_path(ref)
+    if windows_ref is not None:
+        return windows_ref
+
+    try:
+        parsed = urllib.parse.urlsplit(ref)
+    except ValueError:
+        parsed = None
+    if parsed is not None and parsed.scheme:
+        # Do not mistake a Windows drive letter for a URI scheme on a platform
+        # where Windows paths are not native.
+        if WINDOWS_ABSOLUTE_PATH_RE.match(ref):
+            return Path(ref)
+        if parsed.scheme.lower() != "file":
+            return None
+        path_text = urllib.parse.unquote(parsed.path)
+        if parsed.netloc and parsed.netloc.lower() != "localhost":
+            path_text = f"//{parsed.netloc}{path_text}"
+        if os.name == "nt" and re.match(r"^/[A-Za-z]:/", path_text):
+            path_text = path_text[1:]
+        file_path = Path(path_text)
+        return file_path if file_path.is_absolute() else ctx.root / file_path
+
+    if SCHEMELESS_WEB_SOURCE_RE.match(ref):
+        return None
+
+    explicit_relative = ref.startswith(("../", "./", "..\\", ".\\"))
+    home_relative = ref == "~" or ref.startswith(("~/", "~\\"))
+    absolute = Path(ref).is_absolute()
+    unambiguous_bare_relative = (
+        not any(char.isspace() for char in ref) and ("/" in ref or "\\" in ref)
+    )
+    if not (explicit_relative or home_relative or absolute or unambiguous_bare_relative):
+        return None
+
+    ref_path = expand_leading_tilde(ref)
+    if ref_path.is_absolute():
+        return ref_path
+    if explicit_relative:
+        return owner.parent / ref_path
+    return ctx.root / ref_path
+
+
+def source_path_exists(path: Path) -> bool:
+    """Check a source file or directory without crashing on sandbox denials."""
+    try:
+        return path.exists()
+    except OSError:
+        return False
+
+
 def strip_matching_quotes(value: str) -> str:
     if (value.startswith('"') and value.endswith('"')) or (
         value.startswith("'") and value.endswith("'")
diff --git a/tests/fixtures/defects/raw-source-unresolved/_index.md b/tests/fixtures/defects/raw-source-unresolved/_index.md
new file mode 100644
index 0000000..4852262
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/_index.md
@@ -0,0 +1,36 @@
+# Test Wiki
+
+Research on testing patterns for validation purposes.
+
+## Stats
+- Articles: 2
+- Sources: 4
+- Inventory records: 2
+- Datasets: 1
+- Last updated: 2026-01-03
+
+## Articles
+
+| Article | Summary |
+|---------|---------|
+| [Sample Concept](wiki/concepts/sample-concept.md) | A concept about testing patterns synthesized from two sources |
+| [Sample Reference](wiki/references/sample-reference.md) | Reference listing of testing tools and frameworks |
+
+## Outputs
+
+| Output | Type | Date |
+|--------|------|------|
+
+## Raw Sources
+
+See [raw/_index.md](raw/_index.md) for 4 ingested sources.
+
+## Quick Navigation
+
+- [All Sources](raw/_index.md)
+- [Inventory](inventory/_index.md)
+- [Datasets](datasets/_index.md)
+- [Concepts](wiki/concepts/_index.md)
+- [Topics](wiki/topics/_index.md)
+- [References](wiki/references/_index.md)
+- [Outputs](output/_index.md)
diff --git a/tests/fixtures/defects/raw-source-unresolved/config.md b/tests/fixtures/defects/raw-source-unresolved/config.md
new file mode 100644
index 0000000..7a76f2c
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/config.md
@@ -0,0 +1,15 @@
+---
+title: "Test Wiki"
+description: "A minimal wiki fixture for structural validation testing"
+created: 2026-01-01
+---
+
+# Wiki Configuration
+
+## Scope
+
+Testing patterns and frameworks for LLM-powered tools.
+
+## Conventions
+
+Standard wiki conventions apply.
diff --git a/tests/fixtures/defects/raw-source-unresolved/datasets/_index.md b/tests/fixtures/defects/raw-source-unresolved/datasets/_index.md
new file mode 100644
index 0000000..bee14be
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/datasets/_index.md
@@ -0,0 +1,18 @@
+# Dataset Registry Index
+
+> Dataset manifests for large or external data indexed by this wiki.
+
+Last updated: 2026-01-03
+
+## Statistics
+
+- Datasets: 1
+- Active: 0
+- External: 1
+- Unavailable: 0
+
+## Contents
+
+| Dataset | Status | Storage | Formats | Size | Records | Updated |
+|---------|--------|---------|---------|------|---------|---------|
+| [Bitcointalk Temporal Graph](bitcointalk-temporal-graph/MANIFEST.md) | external | external | csv, zip | unknown | unknown | 2026-01-03 |
diff --git a/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/MANIFEST.md b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/MANIFEST.md
new file mode 100644
index 0000000..9addf3b
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/MANIFEST.md
@@ -0,0 +1,48 @@
+---
+title: "Bitcointalk Temporal Graph"
+dataset_id: bitcointalk-temporal-graph
+status: external
+storage: external
+locations:
+  - https://figshare.com/articles/dataset/BitcoinTemporalGraph/26305093
+formats: [csv, zip]
+size_bytes: null
+record_count: null
+schema_status: unknown
+created: 2026-01-03
+updated: 2026-01-03
+tags: [bitcoin, bitcointalk, graph-dataset]
+summary: "A sample external dataset manifest used to validate the dataset registry."
+origin: output/sample-output.md
+inventory:
+  - inventory/candidates/bitcointalk-archive.md
+license: unknown
+access: public
+---
+
+# Bitcointalk Temporal Graph
+
+## Scope
+
+External dataset fixture for validating wiki dataset manifests.
+
+## Storage Locations
+
+The sample manifest points at an external Figshare URL. The fixture does not
+store dataset contents in the wiki.
+
+## Schema
+
+Unknown.
+
+## Samples And Profiles
+
+No samples or profiles are recorded in this fixture.
+
+## Query Recipes
+
+No query recipes are recorded in this fixture.
+
+## Caveats
+
+Fixture data only.
diff --git a/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/_index.md b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/_index.md
new file mode 100644
index 0000000..71c6e40
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/_index.md
@@ -0,0 +1,17 @@
+# Bitcointalk Temporal Graph Dataset Index
+
+> Manifest, samples, profiles, and query recipes for this dataset.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Summary | Tags | Updated |
+|------|---------|------|---------|
+| [MANIFEST.md](MANIFEST.md) | A sample external dataset manifest used to validate the dataset registry. | bitcoin, bitcointalk, graph-dataset | 2026-01-03 |
+
+## Quick Navigation
+
+- [Samples](samples/_index.md)
+- [Profiles](profiles/_index.md)
+- [Queries](queries/_index.md)
diff --git a/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/profiles/_index.md b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/profiles/_index.md
new file mode 100644
index 0000000..08999ad
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/profiles/_index.md
@@ -0,0 +1,10 @@
+# Bitcointalk Temporal Graph Profiles
+
+> Lightweight profiles for this dataset.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Summary | Tags | Updated |
+|------|---------|------|---------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/queries/_index.md b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/queries/_index.md
new file mode 100644
index 0000000..0b94bb4
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/queries/_index.md
@@ -0,0 +1,10 @@
+# Bitcointalk Temporal Graph Queries
+
+> Read-only query recipes for this dataset.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Summary | Tags | Updated |
+|------|---------|------|---------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/samples/_index.md b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/samples/_index.md
new file mode 100644
index 0000000..7d83879
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/datasets/bitcointalk-temporal-graph/samples/_index.md
@@ -0,0 +1,10 @@
+# Bitcointalk Temporal Graph Samples
+
+> Tiny samples or sampling recipes for this dataset.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Summary | Tags | Updated |
+|------|---------|------|---------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/_index.md b/tests/fixtures/defects/raw-source-unresolved/inventory/_index.md
new file mode 100644
index 0000000..228ca52
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/_index.md
@@ -0,0 +1,30 @@
+# Inventory Index
+
+> Durable tracking records for items, candidates, entities, corpora, and watch items.
+
+Last updated: 2026-01-03
+
+## Statistics
+
+- Total records: 2
+- Items: 1
+- Candidates: 1
+- Entities: 0
+- Corpora: 0
+- Active: 0
+- Blocked: 0
+
+## Quick Navigation
+
+- [Items](items/_index.md)
+- [Candidates](candidates/_index.md)
+- [Entities](entities/_index.md)
+- [Corpora](corpora/_index.md)
+- [Views](views/_index.md)
+
+## Contents
+
+| File | Kind | Status | Priority | Next Action | Updated |
+|------|------|--------|----------|-------------|---------|
+| [trx4m-ring-and-pinion.md](items/trx4m-ring-and-pinion.md) | item | proposed | p1 | Decide whether this is the selected drivetrain default. | 2026-01-03 |
+| [bitcointalk-archive.md](candidates/bitcointalk-archive.md) | ingest-candidate | proposed | p1 | Profile available archive formats before ingestion. | 2026-01-03 |
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/candidates/_index.md b/tests/fixtures/defects/raw-source-unresolved/inventory/candidates/_index.md
new file mode 100644
index 0000000..9172a2a
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/candidates/_index.md
@@ -0,0 +1,11 @@
+# Candidates Inventory
+
+> Ingest candidates, open questions, tasks, watch items, and proposed follow-up work.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Kind | Status | Priority | Next Action | Updated |
+|------|------|--------|----------|-------------|---------|
+| [bitcointalk-archive.md](bitcointalk-archive.md) | ingest-candidate | proposed | p1 | Profile available archive formats before ingestion. | 2026-01-03 |
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/candidates/bitcointalk-archive.md b/tests/fixtures/defects/raw-source-unresolved/inventory/candidates/bitcointalk-archive.md
new file mode 100644
index 0000000..425018c
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/candidates/bitcointalk-archive.md
@@ -0,0 +1,33 @@
+---
+title: "Bitcointalk Archive"
+kind: ingest-candidate
+status: proposed
+priority: p1
+created: 2026-01-03
+updated: 2026-01-03
+last_checked: 2026-01-03
+next_action: "Profile available archive formats before ingestion."
+sources:
+  - output/sample-output.md
+tags: [bitcoin, archive, ingest-candidate]
+summary: "A sample ingest candidate used to validate inventory structure."
+---
+
+# Bitcointalk Archive
+
+## Why Track This
+
+This fixture represents a source queue item that should persist without being
+treated as a raw source or compiled wiki article.
+
+## Current State
+
+Proposed for future ingestion.
+
+## Next Action
+
+Profile available archive formats before ingestion.
+
+## Related
+
+- [Sample output](../../output/sample-output.md)
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/corpora/_index.md b/tests/fixtures/defects/raw-source-unresolved/inventory/corpora/_index.md
new file mode 100644
index 0000000..15d2c55
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/corpora/_index.md
@@ -0,0 +1,10 @@
+# Corpora Inventory
+
+> Source collections, archives, datasets, forums, document sets, and other bounded bodies of material.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Kind | Status | Priority | Next Action | Updated |
+|------|------|--------|----------|-------------|---------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/entities/_index.md b/tests/fixtures/defects/raw-source-unresolved/inventory/entities/_index.md
new file mode 100644
index 0000000..7c4c346
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/entities/_index.md
@@ -0,0 +1,10 @@
+# Entities Inventory
+
+> People, organizations, projects, venues, standards bodies, and other named things worth tracking.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Kind | Status | Priority | Next Action | Updated |
+|------|------|--------|----------|-------------|---------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/items/_index.md b/tests/fixtures/defects/raw-source-unresolved/inventory/items/_index.md
new file mode 100644
index 0000000..c6362a8
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/items/_index.md
@@ -0,0 +1,11 @@
+# Items Inventory
+
+> Physical and digital inventory items, parts, tools, assets, hosts, products, and SKUs.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Kind | Status | Priority | Next Action | Updated |
+|------|------|--------|----------|-------------|---------|
+| [trx4m-ring-and-pinion.md](trx4m-ring-and-pinion.md) | item | proposed | p1 | Decide whether this is the selected drivetrain default. | 2026-01-03 |
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/items/trx4m-ring-and-pinion.md b/tests/fixtures/defects/raw-source-unresolved/inventory/items/trx4m-ring-and-pinion.md
new file mode 100644
index 0000000..303f56a
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/items/trx4m-ring-and-pinion.md
@@ -0,0 +1,44 @@
+---
+title: "TRX-4M Ring And Pinion"
+kind: item
+status: proposed
+priority: p1
+created: 2026-01-03
+updated: 2026-01-03
+last_checked: 2026-01-03
+category: drivetrain
+quantity: 1
+unit: set
+state: selected
+default_choice: "Treal hardened helical 12T/24T"
+alternatives:
+  - Traxxas TRA9779
+  - MEUS
+  - LGRP
+  - Injora alloy
+needed_for: trx4m-build
+next_action: "Decide whether this is the selected drivetrain default."
+sources:
+  - wiki/references/sample-reference.md
+tags: [trx4m, parts, drivetrain, item]
+summary: "Sample physical part inventory record used to validate item inventory structure."
+---
+
+# TRX-4M Ring And Pinion
+
+## Why Track This
+
+This fixture represents an actual physical part or SKU that belongs in
+inventory rather than in a generic candidate queue.
+
+## Current State
+
+Proposed as the selected drivetrain default for a TRX-4M build.
+
+## Next Action
+
+Decide whether this is the selected drivetrain default.
+
+## Related
+
+- [Sample reference](../../wiki/references/sample-reference.md)
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/views/_index.md b/tests/fixtures/defects/raw-source-unresolved/inventory/views/_index.md
new file mode 100644
index 0000000..bb985f0
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/views/_index.md
@@ -0,0 +1,11 @@
+# Inventory Views
+
+> Generated inventory views and filtered lists.
+
+Last updated: 2026-01-03
+
+## Contents
+
+| File | Summary | Tags | Updated |
+|------|---------|------|---------|
+| [active-actions.md](active-actions.md) | Derived chat-friendly table of active inventory records with next actions. | inventory, actions, view | 2026-01-03 |
diff --git a/tests/fixtures/defects/raw-source-unresolved/inventory/views/active-actions.md b/tests/fixtures/defects/raw-source-unresolved/inventory/views/active-actions.md
new file mode 100644
index 0000000..8682ab0
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/inventory/views/active-actions.md
@@ -0,0 +1,18 @@
+---
+title: "Active Inventory Actions"
+view: actions
+filters:
+  status: active
+updated: 2026-01-03
+tags: [inventory, actions, view]
+summary: "Derived chat-friendly table of active inventory records with next actions."
+---
+
+# Active Inventory Actions
+
+Generated from inventory record frontmatter on 2026-01-03.
+
+| Record | Kind | Priority | Next Action | Updated |
+|--------|------|----------|-------------|---------|
+
+No active records in this fixture.
diff --git a/tests/fixtures/defects/raw-source-unresolved/log.md b/tests/fixtures/defects/raw-source-unresolved/log.md
new file mode 100644
index 0000000..12ff56c
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/log.md
@@ -0,0 +1,10 @@
+# Wiki Activity Log
+
+## [2026-01-01] init | Wiki initialized
+## [2026-01-01] ingest | Sample Article Source (raw/articles/2026-01-01-sample-article.md)
+## [2026-01-01] ingest | Sample Paper Source (raw/papers/2026-01-01-sample-paper.md)
+## [2026-01-02] ingest | Second Article Source (raw/articles/2026-01-02-second-article.md)
+## [2026-01-03] ingest | Title Cased Source (raw/articles/2026-01-03-Title Cased Source.md)
+## [2026-01-03] compile | 4 sources → 2 articles (sample-concept, sample-reference)
+
+## [2026-01-10] schema | created default advisory schema.md
diff --git a/tests/fixtures/defects/raw-source-unresolved/output/_index.md b/tests/fixtures/defects/raw-source-unresolved/output/_index.md
new file mode 100644
index 0000000..cd6bafc
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/output/_index.md
@@ -0,0 +1,5 @@
+# Output Artifacts
+
+| Output | Type | Date |
+|--------|------|------|
+| [Sample Output](sample-output.md) | report | 2026-01-03 |
diff --git a/tests/fixtures/defects/raw-source-unresolved/output/sample-output.md b/tests/fixtures/defects/raw-source-unresolved/output/sample-output.md
new file mode 100644
index 0000000..7c8a9e1
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/output/sample-output.md
@@ -0,0 +1,12 @@
+---
+title: "Sample Output"
+type: report
+created: 2026-01-03
+updated: 2026-01-03
+summary: "Sample output artifact used by inventory provenance tests."
+---
+
+# Sample Output
+
+This fixture exists so inventory records can point at a generated artifact and
+structural tests can verify local inventory `sources:` paths.
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/_index.md b/tests/fixtures/defects/raw-source-unresolved/raw/_index.md
new file mode 100644
index 0000000..3a93de3
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/_index.md
@@ -0,0 +1,20 @@
+# Raw Sources Index
+
+## Articles
+See [articles/_index.md](articles/_index.md)
+
+## Papers
+See [papers/_index.md](papers/_index.md)
+
+## Repos
+See [repos/_index.md](repos/_index.md)
+
+## Notes
+See [notes/_index.md](notes/_index.md)
+
+## Data
+See [data/_index.md](data/_index.md)
+
+## Stats
+- Total sources: 4
+- Last updated: 2026-01-03
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-01-sample-article.md b/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-01-sample-article.md
new file mode 100644
index 0000000..e07a29a
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-01-sample-article.md
@@ -0,0 +1,14 @@
+---
+title: "Sample Article Source"
+source: https://example.com/testing-patterns
+type: articles
+ingested: 2026-01-01
+tags: [testing, patterns]
+summary: "An article about testing patterns for LLM-powered tools."
+---
+
+# Sample Article Source
+
+Testing LLM-powered tools requires a layered approach: structural validation for file integrity, behavioral evals for instruction compliance, and integration scenarios for full workflows.
+
+The three-layer model is the consensus across Anthropic, OpenAI, and GitLab.
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-02-second-article.md b/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-02-second-article.md
new file mode 100644
index 0000000..292c747
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-02-second-article.md
@@ -0,0 +1,14 @@
+---
+title: "Second Article Source"
+source: https://example.com/eval-frameworks
+type: articles
+ingested: 2026-01-02
+tags: [testing, frameworks]
+summary: "A survey of evaluation frameworks for testing AI agents."
+---
+
+# Second Article Source
+
+Promptfoo and DeepEval are the two leading open-source eval frameworks. Promptfoo is YAML-driven and language-agnostic. DeepEval is Python-native with richer agent metrics.
+
+Both support CI/CD integration via GitHub Actions.
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-03-Title Cased Source.md b/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-03-Title Cased Source.md
new file mode 100644
index 0000000..87d5cf7
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/articles/2026-01-03-Title Cased Source.md	
@@ -0,0 +1,14 @@
+---
+title: "Title Cased Source"
+source: https://example.com/title-cased-source
+type: articles
+ingested: 2026-01-03
+tags: [testing, provenance]
+summary: "A source with whitespace and title case in its filename for resolver regression coverage."
+---
+
+# Title Cased Source
+
+This source intentionally keeps whitespace and title case in the filename. It
+proves source provenance checks preserve complete paths rather than splitting
+on whitespace.
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/articles/_index.md b/tests/fixtures/defects/raw-source-unresolved/raw/articles/_index.md
new file mode 100644
index 0000000..a832a70
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/articles/_index.md
@@ -0,0 +1,7 @@
+# Articles
+
+| Date | Title | Quality | Tags |
+|------|-------|---------|------|
+| 2026-01-01 | [Sample Article Source](2026-01-01-sample-article.md) | 4/5 | testing, patterns |
+| 2026-01-02 | [Second Article Source](2026-01-02-second-article.md) | 4/5 | testing, frameworks |
+| 2026-01-03 | [Title Cased Source](<2026-01-03-Title Cased Source.md>) | 4/5 | testing, provenance |
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/data/_index.md b/tests/fixtures/defects/raw-source-unresolved/raw/data/_index.md
new file mode 100644
index 0000000..db9230f
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/data/_index.md
@@ -0,0 +1,4 @@
+# Data
+
+| Date | Title | Quality | Tags |
+|------|-------|---------|------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/notes/_index.md b/tests/fixtures/defects/raw-source-unresolved/raw/notes/_index.md
new file mode 100644
index 0000000..d0d1e97
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/notes/_index.md
@@ -0,0 +1,4 @@
+# Notes
+
+| Date | Title | Quality | Tags |
+|------|-------|---------|------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/papers/2026-01-01-sample-paper.md b/tests/fixtures/defects/raw-source-unresolved/raw/papers/2026-01-01-sample-paper.md
new file mode 100644
index 0000000..511c31c
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/papers/2026-01-01-sample-paper.md
@@ -0,0 +1,14 @@
+---
+title: "Sample Paper Source"
+source: ../../../research/nonexistent-original.pdf
+type: papers
+ingested: 2026-01-01
+tags: [evals, methodology]
+summary: "A paper on evaluation methodology for AI agents, covering pass@k reliability metrics."
+---
+
+# Sample Paper Source
+
+Pass@k measures capability (at least 1 success in k runs). Pass^k measures reliability (all k runs succeed). A system can score 100% pass@10 and 0% pass^10 simultaneously.
+
+For production systems, pass^k is the standard.
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/papers/_index.md b/tests/fixtures/defects/raw-source-unresolved/raw/papers/_index.md
new file mode 100644
index 0000000..b712b72
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/papers/_index.md
@@ -0,0 +1,5 @@
+# Papers
+
+| Date | Title | Quality | Tags |
+|------|-------|---------|------|
+| 2026-01-01 | [Sample Paper Source](2026-01-01-sample-paper.md) | 5/5 | evals, methodology |
diff --git a/tests/fixtures/defects/raw-source-unresolved/raw/repos/_index.md b/tests/fixtures/defects/raw-source-unresolved/raw/repos/_index.md
new file mode 100644
index 0000000..c0fde83
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/raw/repos/_index.md
@@ -0,0 +1,4 @@
+# Repos
+
+| Date | Title | Quality | Tags |
+|------|-------|---------|------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/schema.md b/tests/fixtures/defects/raw-source-unresolved/schema.md
new file mode 100644
index 0000000..b34bcb4
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/schema.md
@@ -0,0 +1,55 @@
+---
+title: "Test Wiki Topic Guide"
+schema_state: advisory
+created: 2026-01-10
+updated: 2026-01-10
+summary: "Human-owned topic guide for local vocabulary, relationships, source boundaries, and conventions."
+---
+
+# Test Wiki Topic Guide
+
+> This is not a database schema and it does not make existing wiki content invalid.
+> It is a human-owned guide for local vocabulary, relationships, source boundaries, and conventions.
+> The librarian may propose improvements, but agents should not rewrite this file without explicit approval.
+
+## State
+
+- `schema_state`: `advisory`
+- `advisory` means suggestions only.
+- Keep this advisory until the librarian's topic-guide suggestions are consistently low-noise.
+- `strict` is an advanced, explicit opt-in; it still never permits automatic content rewrites.
+
+## Entity Types
+
+| Type | Meaning |
+|------|---------|
+| `concept` | Bounded idea or mechanism compiled under `wiki/concepts/`. |
+| `topic` | Broad theme or playbook compiled under `wiki/topics/`. |
+| `reference` | Curated list, map, standard, or lookup page under `wiki/references/`. |
+| `source` | Raw evidence under `raw/`; factual claims should trace back here. |
+| `artifact` | Generated output, project file, inventory record, or dataset manifest. |
+
+## Relationship Verbs
+
+- `cites`: article or output cites a raw source.
+- `supports`: source or article supports a claim, plan, or decision.
+- `contradicts`: source or article conflicts with another claim.
+- `supersedes`: newer article/output replaces an older one.
+- `depends-on`: artifact or workflow relies on another artifact.
+- `implements`: code, output, or project implements a plan.
+- `relates-to`: weak relationship used when a stronger verb is not yet justified.
+
+## Source Conventions
+
+- Keep raw sources immutable under `raw/`.
+- Compile durable synthesis under `wiki/` with explicit `sources:` frontmatter.
+- Keep generated deliverables under `output/`.
+- Use inventory for durable tracking state, not factual evidence.
+- Use dataset manifests for large, mutable, binary, remote, or query-oriented data.
+
+## Adoption Notes
+
+- Existing articles do not need immediate rewrites to adopt this guide.
+- Librarian topic-guide scans should propose changes in `output/schema-proposal-*.md`.
+- Promote only the small conventions that fit this topic; delete unused starter rows.
+- Prefer staying advisory unless the human explicitly wants stricter convention warnings.
diff --git a/tests/fixtures/defects/raw-source-unresolved/wiki/_index.md b/tests/fixtures/defects/raw-source-unresolved/wiki/_index.md
new file mode 100644
index 0000000..0e65361
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/wiki/_index.md
@@ -0,0 +1,13 @@
+# Wiki Articles
+
+## Concepts
+See [concepts/_index.md](concepts/_index.md)
+
+## Topics
+See [topics/_index.md](topics/_index.md)
+
+## References
+See [references/_index.md](references/_index.md)
+
+## Theses
+See [theses/_index.md](theses/_index.md)
diff --git a/tests/fixtures/defects/raw-source-unresolved/wiki/concepts/_index.md b/tests/fixtures/defects/raw-source-unresolved/wiki/concepts/_index.md
new file mode 100644
index 0000000..763c5f7
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/wiki/concepts/_index.md
@@ -0,0 +1,5 @@
+# Concepts
+
+| Article | Summary | Tags |
+|---------|---------|------|
+| [Sample Concept](sample-concept.md) | Testing patterns for LLM tools — three-layer model | testing, patterns, evals |
diff --git a/tests/fixtures/defects/raw-source-unresolved/wiki/concepts/sample-concept.md b/tests/fixtures/defects/raw-source-unresolved/wiki/concepts/sample-concept.md
new file mode 100644
index 0000000..71f653e
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/wiki/concepts/sample-concept.md
@@ -0,0 +1,31 @@
+---
+title: "Sample Concept"
+category: concept
+sources:
+  - raw/articles/2026-01-01-sample-article.md
+  - raw/papers/2026-01-01-sample-paper.md
+  - "raw/articles/2026-01-03-Title Cased Source.md"
+created: 2026-01-01
+updated: 2026-01-03
+tags: [testing, patterns, evals]
+confidence: high
+volatility: warm
+verified: 2026-01-03
+summary: "Testing patterns for LLM tools — three-layer model with pass@k reliability metrics."
+---
+
+# Sample Concept
+
+Testing LLM-powered tools requires a layered approach. The three-layer model splits tests into structural validation (deterministic, no LLM), behavioral evals (semantic, LLM-graded), and integration scenarios (end-to-end workflows). For framework comparisons, see [[sample-reference|Sample Reference]] ([Sample Reference](../references/sample-reference.md)).
+
+Pass@k measures capability while pass^k measures reliability. Both metrics are needed.
+
+## See Also
+
+- [[sample-reference|Sample Reference]] ([Sample Reference](../references/sample-reference.md)) — tools and frameworks for implementing tests
+
+## Sources
+
+- [Sample Article Source](../../raw/articles/2026-01-01-sample-article.md) — three-layer testing model
+- [Sample Paper Source](../../raw/papers/2026-01-01-sample-paper.md) — pass@k reliability metrics
+- [Title Cased Source](<../../raw/articles/2026-01-03-Title Cased Source.md>) — whitespace filename source-resolution regression
diff --git a/tests/fixtures/defects/raw-source-unresolved/wiki/references/_index.md b/tests/fixtures/defects/raw-source-unresolved/wiki/references/_index.md
new file mode 100644
index 0000000..a3c5b57
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/wiki/references/_index.md
@@ -0,0 +1,5 @@
+# References
+
+| Article | Summary | Tags |
+|---------|---------|------|
+| [Sample Reference](sample-reference.md) | Eval frameworks — Promptfoo, DeepEval, comparison | testing, frameworks, tools |
diff --git a/tests/fixtures/defects/raw-source-unresolved/wiki/references/sample-reference.md b/tests/fixtures/defects/raw-source-unresolved/wiki/references/sample-reference.md
new file mode 100644
index 0000000..c961d2d
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/wiki/references/sample-reference.md
@@ -0,0 +1,31 @@
+---
+title: "Sample Reference"
+category: reference
+sources:
+  - raw/articles/2026-01-02-second-article.md
+created: 2026-01-02
+updated: 2026-01-02
+tags: [testing, frameworks, tools]
+confidence: high
+volatility: warm
+verified: 2026-01-02
+summary: "Eval frameworks — Promptfoo (YAML, language-agnostic) vs DeepEval (Python, richer metrics)."
+---
+
+# Sample Reference
+
+## Promptfoo
+
+Open-source, YAML-driven, CLI-first eval framework with native Claude Agent SDK support.
+
+## DeepEval
+
+Python-native, pytest-based. Eight agent-specific metrics including Task Completion and Tool Correctness.
+
+## See Also
+
+- [[sample-concept|Sample Concept]] ([Sample Concept](../concepts/sample-concept.md)) — the testing methodology these tools implement
+
+## Sources
+
+- [Second Article Source](../../raw/articles/2026-01-02-second-article.md) — framework comparison
diff --git a/tests/fixtures/defects/raw-source-unresolved/wiki/theses/_index.md b/tests/fixtures/defects/raw-source-unresolved/wiki/theses/_index.md
new file mode 100644
index 0000000..601ebb2
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/wiki/theses/_index.md
@@ -0,0 +1,4 @@
+# Theses
+
+| Thesis | Status | Verdict | Tags |
+|--------|--------|---------|------|
diff --git a/tests/fixtures/defects/raw-source-unresolved/wiki/topics/_index.md b/tests/fixtures/defects/raw-source-unresolved/wiki/topics/_index.md
new file mode 100644
index 0000000..23e91d2
--- /dev/null
+++ b/tests/fixtures/defects/raw-source-unresolved/wiki/topics/_index.md
@@ -0,0 +1,4 @@
+# Topics
+
+| Article | Summary | Tags |
+|---------|---------|------|
diff --git a/tests/generate-defect-fixtures.sh b/tests/generate-defect-fixtures.sh
index 90c2577..a08cdf2 100755
--- a/tests/generate-defect-fixtures.sh
+++ b/tests/generate-defect-fixtures.sh
@@ -75,6 +75,13 @@ rm -f "$DEFECTS/dangling-source-ref/wiki/concepts/sample-concept.md.bak" \
   "$DEFECTS/dangling-source-ref/inventory/items/trx4m-ring-and-pinion.md.bak"
 echo "  Created: dangling-source-ref (C4b)"
 
+# C4b: raw-source-unresolved — explicit local source: path does not resolve
+copy_golden "raw-source-unresolved"
+sed -i.bak 's|^source: https://example.com/eval-methodology$|source: ../../../research/nonexistent-original.pdf|' \
+  "$DEFECTS/raw-source-unresolved/raw/papers/2026-01-01-sample-paper.md"
+rm -f "$DEFECTS/raw-source-unresolved/raw/papers/2026-01-01-sample-paper.md.bak"
+echo "  Created: raw-source-unresolved (C4b)"
+
 # C4b: retracted-marker — <!--RETRACTED-SOURCE--> left in body
 copy_golden "retracted-marker"
 echo '<!--RETRACTED-SOURCE: previously cited claim from deleted source-->' \
diff --git a/tests/test-local-cli-lint.sh b/tests/test-local-cli-lint.sh
index 441e51c..ddbf647 100755
--- a/tests/test-local-cli-lint.sh
+++ b/tests/test-local-cli-lint.sh
@@ -284,6 +284,34 @@ expect_failure_contains \
   "Invalid type" \
   "$CLI" lint "$SCRIPT_DIR/fixtures/defects/bad-frontmatter"
 
+expect_failure_contains \
+  "explicit unresolved raw source path fails local lint" \
+  "Raw source reference does not resolve" \
+  "$CLI" lint "$SCRIPT_DIR/fixtures/defects/raw-source-unresolved"
+
+raw_source_compat="$tmpdir/raw-source-compat"
+mkdir "$raw_source_compat"
+cp -R "$GOLDEN/." "$raw_source_compat/"
+mkdir "$tmpdir/local-source-repository"
+mkdir "$tmpdir/local source files"
+printf '# Original source\n' > "$tmpdir/local source files/original.md"
+sed -i.bak 's|^source: https://example.com/testing-patterns$|source: session|' \
+  "$raw_source_compat/raw/articles/2026-01-01-sample-article.md"
+sed -i.bak 's|^source: https://example.com/eval-frameworks$|source: s3://example-bucket/source.md|' \
+  "$raw_source_compat/raw/articles/2026-01-02-second-article.md"
+sed -i.bak "s|^source: https://example.com/title-cased-source$|source: ../../../local source files/original.md|" \
+  "$raw_source_compat/raw/articles/2026-01-03-Title Cased Source.md"
+sed -i.bak "s|^source: https://example.com/eval-methodology$|source: file://$tmpdir/local-source-repository|" \
+  "$raw_source_compat/raw/papers/2026-01-01-sample-paper.md"
+rm -f "$raw_source_compat"/raw/articles/*.bak "$raw_source_compat"/raw/papers/*.bak
+cat >> "$raw_source_compat/raw/articles/2026-01-01-sample-article.md" <<'EOF'
+
+Upstream navigation remains source content: [contributor guide](docs/CONTRIBUTING.md)
+EOF
+expect_success \
+  "raw provenance sentinels, URIs, valid paths, directories, and upstream links stay compatible" \
+  "$CLI" lint "$raw_source_compat"
+
 ideas_wiki="$tmpdir/ideas-wiki"
 mkdir "$ideas_wiki"
 cp -R "$GOLDEN/." "$ideas_wiki/"
diff --git a/tests/test-structure.sh b/tests/test-structure.sh
index 9f5227a..bb090e3 100755
--- a/tests/test-structure.sh
+++ b/tests/test-structure.sh
@@ -452,6 +452,12 @@ if [ -d "$DEFECTS" ]; then
       || log_fail "dangling-source-ref: no dangling ref" "fixture broken"
   }
 
+  [ -d "$DEFECTS/raw-source-unresolved" ] && {
+    grep -q "source: ../../../research/nonexistent-original.pdf" "$DEFECTS/raw-source-unresolved/raw/papers/2026-01-01-sample-paper.md" 2>/dev/null \
+      && log_pass "raw-source-unresolved: C4b defect present" \
+      || log_fail "raw-source-unresolved: no unresolved raw source path" "fixture broken"
+  }
+
   [ -d "$DEFECTS/retracted-marker" ] && {
     grep -q "RETRACTED-SOURCE" "$DEFECTS/retracted-marker/wiki/concepts/sample-concept.md" 2>/dev/null \
       && log_pass "retracted-marker: C4b defect present" \