[andy-zen-dev]

Closes #269.

Product outcome

python -m scriptorium.analyze manuscript.txt --output analysis.json

python -m scriptorium.analyze manuscript.txt --format html --output report.html

Uses the existing deterministic engine directly: all 29 rows, units, profiles, dictionary identity and raw/normalized hashes. Optional UTF-8 dictionary/profile pairing; explicit short-input and unavailable-value limitations; no manuscript/dictionary contents or external services. Accessible offline HTML includes all metric families and profile details. Output publication is atomic and create-only, including existing files, aliases and dangling symlinks. No metric formulas or scientific gates change.

Verification at 752f3c4

• 22 focused tests PASS against byte-identical current core files (metrics/text/dialogue/vocabulary), not behavior-equivalent substitutes.
• Actual module CLI: JSON and HTML each rebuilt twice with byte-identical output.
• Engine-value equality, Unicode/raw identity, dictionary privacy, empty input, threshold boundaries, failed writes and output collision tests PASS.
• Full repository CI: pending hosted exact-head execution.
• Browser visual verification: NOT_RUN at PR creation; structural keyboard/table checks PASS.

Scope / integration

Independent of #268; only an additive README section overlaps that PR and it is at a different location. Adopts and hardens the pending local-analysis handoff rather than maintaining a parallel implementation. No Pages activation or package-registry publication. Later fresh acceptance is required before merge; this turn authors the change.

[andy-zen-dev]

Exact-head verification completed for 752f3c491af2f1c563a0c5ae31e9452f93a1c0a3:

• All five returned applicable PR-triggered workflow runs completed success.
• Pages run 37618413182, job 112782285170: full standard-library test suite, canonical site build, deterministic double build and artifact upload all passed. Deploy correctly skipped on this PR.
• The 22 focused local tests use byte-identical canonical metrics/text/dialogue/vocabulary modules and exercise the real engine, not substitutes. Actual CLI JSON and HTML rebuilds were byte-identical. Committed scriptorium/analyze.py blob b99886b5ca46e175efc4c70b39b7b81408b12cd3 matches the tested implementation.
• Full-suite execution is now verified in hosted CI; the earlier PR-body pending status was at creation. Visual verification of this standalone local-analysis HTML remains NOT_RUN; structural keyboard/table/privacy tests passed.

No merge in this authoring run. This PR is independent of #268/#272 and is eligible for its own later fresh exact-head acceptance without waiting for that stack.

[andy-zen-dev]

Owner-requested review, exact head 752f3c4.

Two reproduced findings in scriptorium/analyze.py (blob b99886b5ca46e175efc4c70b39b7b81408b12cd3):

• P1: main() calls Path.read_bytes()/read_text() before checking file type or size. A local FIFO with no writer blocks indefinitely; reproduced in a child process terminated after a 2-second deadline. Reject special files before blocking reads, bound manuscript/dictionary input, and test the actual CLI.
• P2: parser.error(str(exc)) discloses absolute input/output paths. Reproduced on an existing report; the old report was preserved, but stderr exposed its full path. Return actionable path-free errors.

The analyzer/output separation, unchanged metric engine, null handling and create-only writer are sound in the inspected scope. This is COMMENT-only solo-maintainer review, not GitHub self-approval or merge acceptance. The owner now requests one new integration PR against master; carry the existing implementation and these fixes into that delivery path, retaining #269 and this PR as provenance until the integration is accepted. No scientific gate or formula change is requested.

[andy-zen-dev]

Superseded delivery path: #276, per the owner's 2026-10-18 request for one new PR covering remaining ready issues. The analyzer was adopted through #274 and retained in #276 with the FIFO/non-regular input and private-error fixes plus their regressions; no parallel reimplementation. #276 at d81aa3a46c0125561a1ba610d51a6cf9f06df0c4 now has 23/23 successful PR workflows and a 759-test canonical suite, installed-wheel smoke and deterministic site/gallery builds.

This older head 752f3c491af2f1c563a0c5ae31e9452f93a1c0a3 is unchanged and still has the recorded review findings; do not merge it independently. Converting to draft to make #276 the single active review path. Preserve this PR/history until later successor acceptance confirms issue #269 is delivered; no branch deletion or premature completion.

[andy-zen-dev]

Superseded by merged #276 (merge commit 03d122aaf676385ee91e86e5b4cbdf3d212ab752). Its six historical changed paths are included in merged #276; #270's FIFO/nonblocking-read and path-disclosure findings were fixed there and exact-head CI passed. Original PR commit remains divergent and must not be independently merged.

Closing this stale Draft PR without deleting its branch, force-pushing or claiming that its own old checks now pass. The canonical follow-up belongs to the current master / new feature issues.