Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@governance_blueprint/validation/generate_artifact_manifest.py`:
- Around line 166-170: The current logic indents the full-manifest comparison
under the artifacts branch so that generated_utc/external_artifacts drift is
ignored when artifacts match; in generate_artifact_manifest.py, dedent the
second if (the check using current_obj != expected_obj) so it runs at the same
level as the artifacts comparison (i.e., check artifacts with current_artifacts
and expected_artifacts, then separately check the full manifest using
current_obj and expected_obj) and keep the existing print message when the
full-manifest mismatch is detected.
- Around line 130-133: In build_manifest(), remove the duplicate stale entry
"version": "1.4.5" so the dict only contains the intended "version": "1.4.0" (or
update the single version key to the correct current version) — locate the
returned dict in function build_manifest() and delete the extra "version" line
to prevent the wrong value from being retained by the literal.

In `@governance_blueprint/validation/run_validation_suite.py`:
- Around line 81-86: The code adds three explicit selftest scripts then also
appends everything returned by _selftest_scripts(), causing duplicates; modify
the block where skip_selftest is handled to avoid scheduling duplicates by
either (A) removing the three hardcoded steps and rely solely on
_selftest_scripts(), or (B) filter _selftest_scripts() so you skip any script
equal to "governance_blueprint/validation/selftest_validate_artifacts.py",
"governance_blueprint/validation/selftest_generate_artifact_manifest.py", or
"governance_blueprint/validation/selftest_run_validation_suite.py" before
appending to steps; update the logic around skip_selftest, steps, and
_selftest_scripts() accordingly so each selftest appears only once.

In `@governance_blueprint/validation/selftest_generate_artifact_manifest.py`:
- Around line 50-55: The test is patching a non-existent attribute DEFAULT_FILES
and raising AttributeError; update the test to patch gm.BASE_DEFAULT_FILES
instead wherever DEFAULT_FILES is set/restored (e.g., replace gm.DEFAULT_FILES =
["dummy.txt"] and old_default_files = gm.DEFAULT_FILES with
gm.BASE_DEFAULT_FILES = ["dummy.txt"] and old_base_default_files =
gm.BASE_DEFAULT_FILES, and restore gm.BASE_DEFAULT_FILES from the saved old
value after the test), applying the same change at the other occurrences noted
(lines ~61-63, ~123-128, ~138-140) so the helpers can access the correct
attribute during gm.main().
- Around line 17-18: Move the second "from __future__ import annotations" and
any duplicated module docstring/imports to the top of
governance_blueprint/validation/selftest_generate_artifact_manifest.py (before
DUMMY_CONTENT/DUMMY_HASH) so all future imports are at file-start, and remove
the duplicate imports/docstring; ensure DUMMY_CONTENT and DUMMY_HASH remain
below those top imports. In
governance_blueprint/validation/generate_artifact_manifest.py add a proper
indented body for the "if current_artifacts != expected_artifacts:" check (e.g.,
raise AssertionError or log and return/exit) so the module can import, and
ensure the check performs the intended --check behavior. Finally, update the
test patching approach: instead of patching gm.DEFAULT_FILES (which isn't used),
patch BASE_DEFAULT_FILES or call/override _default_files() in
generate_artifact_manifest so the generator actually sees the intended defaults.

In `@governance_blueprint/validation/validate_artifacts.py`:
- Around line 396-412: The function validate_master_reference_markdown currently
appends missing-section messages to errors but never returns errors on the
normal path; add a final return errors at the end of
validate_master_reference_markdown so it returns [] for successful validations
and the populated list when failures are found (preserve the existing
early-return for the missing file case and keep the required_patterns loop and
errors variable as-is).
- Around line 258-300: The function validate_roadmap_2035_shape() collects
validation messages into the local variable errors but never returns it, causing
callers expecting a list to receive None; fix by adding a final "return errors"
at the end of validate_roadmap_2035_shape so the function's declared return type
list[str] is satisfied and downstream callers (e.g., run_checks() / selftests)
receive the collected errors.

In `@Makefile`:
- Around line 67-70: The Makefile contains a duplicate governance-check target
that merged stale prerequisites (governance-validate and
governance-reports-validate) causing incorrect behavior; remove the
stale/duplicate governance-check definition and keep a single governance-check
target whose prerequisites use the current names (e.g., governance-test,
governance-schema-validate, governance-validate-json-check) and deduplicate the
.PHONY list so only one .PHONY declaration includes governance-check.

---

Outside diff comments:
In `@governance_blueprint/artifact_manifest.json`:
- Around line 3-32: The manifest contains duplicated top-level keys ("version",
"generated_utc") and repeated artifact entries (e.g.,
"validation/validate_artifacts.py", "validation/generate_artifact_manifest.py",
"validation/run_validation_suite.py", "validation/selftest_*" etc.) so JSON
consumers keep only the last occurrence; remove the older duplicate blocks and
keep the intended/current values and hashes (or merge them) so each artifact
path and each top-level key appears exactly once; update the "version" and
"generated_utc" to the correct final values and ensure each validation/* entry
uses the correct, non-duplicated checksum listed in the intended set.