[datacas]

Closes #372 — #372

Summary

The criteria parser now treats a punctuated bold label followed by prose, such as **Tests.** text, as a section boundary. Flat-list validation remains strict. Regression coverage includes the existing standalone label and heading forms, malformed and placeholder criteria, bold markup within criteria, and the unchanged body of #330.

Diagnosis

Issue #372 reports that roadmap criteria are rejected when the next section uses an inline bold label. Reproduction before the fix: read_criteria("github", "datacas/code-cycle-toolkit", "330") returned needs_refinement: criteria must be a flat Markdown list. The cause matched the issue hypothesis: _SECTION recognized only headings and standalone bold labels. The invariant is that every non-section line inside the criteria block must be a flat Markdown list item.

The related-item search was widened because this is a bug in a shared parser. It found #159 and closed issue #253 as search candidates; neither shares this parser cause. #330 is the real-world body fixture. Classification: isolated_defect; decision: implement.

Work units

Unit	Behaviour	Files	Tests	Verification	Rollback
WU-1	Recognize punctuated inline bold section labels while preserving criteria validation.	scripts/verification_plan.py, tests/test_verification_plan.py, tests/fixtures/issue-330.md, docs/verification.md	CriteriaTests regression coverage	Parser smoke passed; python3 scripts/validate-package.py passed; configured full suite exceeded the 10-minute stage limit and is unverified.	Independent

Verification (agent-reported)

• python3 -m unittest discover -s tests — timed out after the 10-minute limit and was stopped (exit 130); not verified on the final head.
• python3 scripts/validate-package.py — passed (exit 0): 16 skills validated and shared sections/manifests aligned.
• python3 -m unittest — exited 0 but ran 0 tests; this is not evidence for the acceptance criteria.
• Parser smoke assertions — passed (exit 0) for all six acceptance criteria against the final worktree.
• Boundary verification: not required; the change is confined to pure parsing behavior.

The local run uses the repository checkout and dispatch environment; CI will run on GitHub's hosted runners with its own environment. Code-host checks will be reported for the pushed head once read.

[datacas]

Implementation is published on task-372. Verification is blocked because the required full test suite did not finish within the ten-minute stage limit; the parser smoke assertions and package validation passed, but they do not replace that required check.

[datacas]

[CCR-20260930-001] · reviewer · anthropic/claude-sonnet-5-5→? · high · schema:2

Initial review: PR #374 (work item #372)

Status: REVIEW_PASSED_AWAITING_CI — head bc026627e7765612ec508ba1cb98ffcdae09b572 fully reviewed; no open finding blocks approval. This is not an approval: the required checks of this head are still running and the cycle's CI gate decides.

Scope

Accumulated diff origin/main...HEAD (119 lines, 4 files): scripts/verification_plan.py, tests/test_verification_plan.py, tests/fixtures/issue-330.md, docs/verification.md. Repository conventions applied: AGENTS.md (docs page updated with the behavioural change: satisfied).

Findings

None. _SECTION gains one alternative, \*\*[^*]+[.:]\*\*[:.]?\s+\S.*$. It requires punctuation inside the bold label plus same-line prose, so:

• **Prose** is still prose. (no punctuation in the label) stays a flat-list violation;
• - Keeps **bold markup** ... items start with a list marker and never match;
• the pre-existing standalone-label and heading forms are untouched;
• nested items and placeholders still return needs_refinement.

Acceptance criteria (tests.version=2)

AC	Check	Result
AC-1	test_inline_bold_section_label_ends_criteria (**Tests.** and **Tests:**)	passed
AC-2	existing standalone-label and heading tests in CriteriaTests	passed
AC-3	test_malformed_criteria_remain_needs_refinement (prose, **Prose** line, nested item)	passed
AC-4	same test (placeholder TODO, empty list)	passed
AC-5	test_bold_markup_inside_criterion_does_not_end_section	passed
AC-6	test_issue_330_body_is_evaluable_without_remote_edit (8 items, fixture of the #330 body)	passed

The AC-6 fixture is a copy of the issue body; it does not read the live issue.

Security triage

Not a sensitive change: the security audit was skipped after triage. The deterministic rule (security_required=false in the plan) did not match any changed path (docs/, scripts/verification_plan.py, tests/), and my own reading finds a pure text parser with no new input surface, no I/O or privilege change.

Commands executed

• python3 -m unittest tests.test_verification_plan — exit 0, 28 tests OK.
• python3 scripts/validate-package.py — exit 0: VALID: 16 skills.
• python3 -m unittest discover -s tests — not verified: did not finish within the 10-minute limit (same as the implementer's run); output up to that point showed only passing dots. Not counted as passed.
• python3 -m unittest (no discovery) runs 0 tests and is no evidence.

Unverified

Full test suite on this head (timeout, see above); CI on GitHub's runners is the remaining evidence.

CI (parallel policy, not waited on)

Workflow "Validate package", all IN_PROGRESS at read time: Skills, manifests, and Bash installer; Windows test shard (windows-1..4). Nothing passed, failed or skipped yet.

Residual risks

A prose line in the criteria list that begins with a punctuated bold label (**Note.** ...) now silently ends the section instead of being rejected. This is the intended grammar and is documented in docs/verification.md.

Verdict

No blocking findings. Pending only the CI gate for the head above.