tools: the PI manifest declared contract evidence it never stat'd

[mdheller]

Closes the declared-but-never-stat'd gap in
tools/validate_professional_intelligence_manifest.py, reported but not fixed
in #1096.

require_set_contains() proves the manifest lists a contract path. Nothing
ever checked the path was there. Deleting all five in-repo evidence files
and re-running the tool on unmodified origin/main, in a temp copy of the repo:

OK: professional-intelligence.manifest.yaml structure valid
OK: workspaceOS contract-aligned evidence present
OK: workroom update runtime boundary refs present
OK: workroom update invalid fixture is registered
OK: claim boundaries present
rc=0

A check that verifies a declaration about the world rather than the world.

What changed

require_declared_paths_exist() stats the in-repo members of
REQUIRED_WORKSPACE_OS_CONTRACTS / REQUIRED_WORKSPACE_OS_CONTROLS. It is
driven off the same constants the listing checks use, so the two cannot
drift, and the manifest cannot weaken the check by editing itself.

Cross-repo refs are skipped out loud

SocioProphet/<repo>:<path> refs resolve into peer repos absent from this
checkout and genuinely cannot be stat'd here. They are skipped deliberately and
the skip is printed, on every run, including green ones:

SKIP: workspaceOS.contractPaths: 3 cross-repo ref(s) not stat-able from this checkout, not verified here: [...]
OK: workspaceOS.contractPaths: 3 in-repo evidence path(s) exist on disk
SKIP: workspaceOS.controlRefs: 4 cross-repo ref(s) not stat-able from this checkout, not verified here: [...]
OK: workspaceOS.controlRefs: 2 in-repo evidence path(s) exist on disk

A silently-skipped ref is how this class comes back. 7 skipped, 5 verified,
both counts visible.

Classification matches <org>/<repo>: against a known-org set, not any
<a>/<b>: shape, so an unrecognised prefix falls through to the LOCAL side and
fails loudly. Skipping is the costly direction of a misclassification. This is
not theoretical: the first draft of the regex sent docs/notes:draft.md to the
skipped column and the new test caught it before commit.

Teeth — the mandatory proof

Each of the five deleted in turn in a temp copy of the repo:

tool	rc	behaviour
origin/main	0	prints OK: workspaceOS contract-aligned evidence present
this branch	2	ERR: workspaceOS.controlRefs declares 'docs/WORKROOM_UPDATE_RUNTIME_BOUNDARY.md' but that path does not exist in this repo

Restored → rc=0 again. All five caught, identical under python3 and
python3 -O. Every one of the 21 new tests runs under both interpreters,
extending the -O discipline #1096 introduced rather than regressing it.

The baseline test asserts the evidence line is present, not merely rc=0 —
rc=0 alone would still pass if the check were deleted.

2. demoAcceptance named the wrong field

A non-mapping demoAcceptance collapsed to None through an inline ternary and
was then reported as "demoAcceptance.required must be a non-empty list". It
fails closed either way (rc=2), so this is diagnostics only. It now routes
through expect_mapping() like every other field in the file:

manifest	before	after
demoAcceptance: [...]	ERR: demoAcceptance.required must be a non-empty list	ERR: demoAcceptance must be a mapping/object
demoAcceptance: {required: []}	ERR: demoAcceptance.required ...	unchanged

Verification

• tools/tests: 310 passed (289 on this base + 21 new). No regressions.
• Real manifest still validates rc=0 under python3, python3 -O, and
  make validate-professional-intelligence-manifest.

⚠️ Merge-order note — interacts with #1096

#1096 is not merged, so this branch is cut from main and its 289-test
baseline excludes #1096's 9 tests. The two merge cleanly textually, but the
combination is not green, and that is worth catching before it lands:

#1096's run_against() builds a fixture root holding only the tool and the
manifest. Under this branch's tool that root is no longer a valid repo — the
five declared evidence paths are absent — so
test_real_manifest_still_validates fails (2 tests; the malformed-shape tests
still pass). The tool is behaving correctly; the fixture is under-specified.

Verified fix, 7 lines in #1096's file — whichever PR merges second applies it:

DECLARED_EVIDENCE = [
    "contracts/workspace/workroom-update-request.example.json",
    "contracts/workspace/workroom-update-response.accepted.example.json",
    "contracts/workspace/workroom-update-response.invalid-runtime-mutation.example.json",
    "docs/WORKROOM_UPDATE_RUNTIME_BOUNDARY.md",
    "tools/validate_workroom_update_contract.py",
]
# ...in run_against(), after copying the tool:
for rel in DECLARED_EVIDENCE:
    dest = root / rel
    dest.parent.mkdir(parents=True, exist_ok=True)
    shutil.copy2(REPO_ROOT / rel, dest)

Measured on the actual merge tree: 30 passed across both files with that
patch applied. Not applied here — the file does not exist on this branch's base.

Lane

tools/ only — two files. Does not touch
.github/workflows/validate-target-diagnostics.yml.

[Copilot]

Pull request overview

Note

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Tightens the professional-intelligence manifest validator so it verifies that declared in-repo evidence paths actually exist, and improves diagnostics for malformed demoAcceptance.

Changes:

• Add an existence check that stats declared in-repo evidence paths while explicitly reporting skipped cross-repo refs.
• Fix demoAcceptance shape validation to report the correct field when it’s not a mapping.
• Add tests that prove the new existence check runs (including under python -O) and fails loudly per missing path.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
tools/validate_professional_intelligence_manifest.py	Adds cross-repo ref classification + declared-path existence verification; improves demoAcceptance diagnostics.
tools/tests/test_professional_intelligence_manifest_evidence_paths.py	Adds regression tests ensuring declared evidence is verified on disk and cross-repo skips are reported.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.