asciidoc-pubkit 0.8.2 → 0.8.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 3dd09ef810a52e64bd286b34f4f59cda2cf357214abd94a1255ee44d77a96e27
-  data.tar.gz: fec7d11ca4f985b41aa1d2e5d37c8b126d56a1dc1ee86874330d2239634207c1
+  metadata.gz: 7f629e73e42aa273d3f13a77a917208bf9d54a7c53f024c89c0f7f8fae596934
+  data.tar.gz: fdfd437f9a8c7735b873b3df5da5c55662be58a7871e5c3017f360f09539c382
 SHA512:
-  metadata.gz: e56e14f6abd189a621a3f4b918d9e24697749bc1a777816b2cdfa9a02bfc662f1b53e831e291af99b5254ac976a6fa72a2703c224d98ce979e0d9d7eac5be6d5
-  data.tar.gz: 19361ba968e9738a736f4452eb6590b90cbc1ca998e128d0cdd952f1da527c66d97817f6913f9957092c8dfbae55b5019e90909e1709337671bb57384031da06
+  metadata.gz: d02a86fc684bed05d233025281de87ecd8adf45ed15ae72b57b6e15c12d3572df48b3697294768a3aab3de5e40e1d02ea0828c5a39efd6c3b8f4d68c75f3f2ee
+  data.tar.gz: 54d8f0862dc0b45f78184613e12098569486541378b0fd72b2e218b0daf94846c3513536620c5a5a6a90197d8f07ef6e1526b153523e3d87e2867dd1c15ac58d

data/CHANGELOG.md

@@ -1,5 +1,29 @@
 # Changelog
 
+## 0.8.4 — 2026-10-02
+
+- Require passage-specific review decisions in prose and heading prompts instead
+  of inferring keep/revise from retained strings. Leave unreviewed items pending,
+  reconcile decisions with edits or proposals, and report mechanical verification,
+  candidate handling, and editorial completion separately. Protect session files
+  and baseline snapshots from editing and formatting tools. These instructions
+  do not add automatic decision-record or semantic validation to verification.
+- Detect inflected invocation, interpretation, retrieval, and collection verbs
+  for varied objects through configured MeCab lemmas; retain contextual phrase
+  precedence, allow lists, and explicit literal-mode coverage. Add the exact
+  contextual phrase 木全体 to cover IPADIC surname segmentation.
+- Add regression cases based on investigated manuscript variants and test review
+  instructions in both scopes and modes with session artifacts kept intact.
+
+## 0.8.3 — 2026-10-02
+
+- Refine shared Japanese writing and review criteria for established technical
+  terminology, precise operations, direct predicates, and evidence-based tone.
+  Preserve distinctions between assumptions, requirements, and capabilities.
+- Add scoped review candidates for unusual translations, everyday operation
+  wording, indirect predicates, and author-centered emphasis; reuse existing
+  noun, verb, and qualification rules instead of fixed automatic replacements.
+
 ## 0.8.2 — 2026-10-02
 
 - Report replacement rule errors with the original YAML filename and one-based

data/README.md

@@ -7,7 +7,7 @@
 
 A toolkit for authoring and reviewing AsciiDoc books.
 
-Version 0.8.2 provides shared Japanese technical writing criteria, an authoring
+Version 0.8.4 provides shared Japanese technical writing criteria, an authoring
 prompt, separate prose and heading review prompts, manuscript scoring, and
 baseline verification for edited manuscripts, plus explicit prose-only mechanical
 replacements using a limited prh-format rule set.
@@ -16,7 +16,7 @@ Japanese text is retained in manuscript excerpts, rule dictionaries, and fixture
 
 ## Available features
 
-The following commands are available in version 0.8.2.
+The following commands are available in version 0.8.4.
 
 | Feature | Commands | Behavior |
 | --- | --- | --- |
@@ -107,7 +107,7 @@ Add the gem to your project's `Gemfile` to manage its version with Bundler:
 
 ```ruby
 source 'https://rubygems.org'
-gem 'asciidoc-pubkit', '~> 0.8.2'
+gem 'asciidoc-pubkit', '~> 0.8.4'
 ```
 
 Then install dependencies and run the CLI through Bundler:
@@ -129,7 +129,7 @@ git clone https://github.com/cybergarage/asciidoc-pubkit.git
 cd asciidoc-pubkit
 bundle install
 gem build asciidoc-pubkit.gemspec
-gem install ./asciidoc-pubkit-0.8.2.gem
+gem install ./asciidoc-pubkit-0.8.4.gem
 asciidoc-pubkit --version
 ```
 
@@ -153,7 +153,7 @@ asciidoc-pubkit writing prompt --lang ja --output writing-prompt.md
 ```
 
 For example, `--lang en` exits with an unsupported-language error. No English
-review or writing criteria are shipped in 0.8.2.
+review or writing criteria are shipped in 0.8.4.
 
 For a small trial, use `examples/book.adoc` as the scan input. Its Japanese
 paragraphs deliberately contain review candidates; its code block must remain
@@ -612,7 +612,21 @@ asciidoc-pubkit review prompt .pubkit/review --mode diagnose
 
 `revise` is the default and asks the agent to edit relevant prose. `diagnose` asks
 for findings and proposed revisions without editing. Both modes require the agent
-to distinguish **revise**, **keep**, and **needs-evidence** decisions.
+to distinguish **revise**, **keep**, and **needs-evidence** decisions. Each decision
+needs a passage-specific reason. Whether a matched string remains or disappears
+cannot determine that decision; unchanged paragraphs and remaining candidates
+must not receive automatic `keep` records. Unreviewed items remain pending.
+
+Version 0.8.4 prompts require the reviewer to reconcile decisions with actual
+edits (or proposed revisions in diagnose mode), review paragraphs or headings
+without candidates, and inspect newly detected or remaining candidates after
+verification. They request separate reports for mechanical preservation,
+candidate decisions, and paragraph or heading quality review, with counts and
+unresolved items. Protected-scope concerns are separate proposals, not evidence
+of editorial approval. Pending items make the review incomplete.
+These are reviewer instructions: the gem does not validate a decision ledger
+or prove that an AI or human followed the instructions. Keep session metadata
+and baseline snapshots immutable; format only authorized manuscript spans.
 
 The single Markdown output includes every selected paragraph once, in document
 order, with the shared criteria saved at scan time, candidates, saved settings,
@@ -649,7 +663,10 @@ Non-prose comparison ignores empty separator lines and trailing whitespace;
 listing, literal, and passthrough block content is additionally compared as parsed
 lines. Inline protection is heuristic, not a complete AsciiDoc inline parser.
 Successful verification does not prove meaning preservation, technical accuracy,
-or release readiness: `meaning_verified` is always `false`.
+or release readiness: `meaning_verified` is always `false`. Verification does
+not read candidate decision records or certify editorial completion. A passing
+mechanical result must be reported separately from the reviewer's quality and
+completion assessment.
 
 | Exit code | Meaning |
 | --- | --- |
@@ -1005,6 +1022,28 @@ These are candidates for contextual review, not banned expressions. A necessary
 condition, uncertainty, or distinction must survive a revision; a redundant
 disclaimer can instead be removed or folded into a more precise main claim.
 
+Version 0.8.3 also includes scoped candidates such as `構築入口`,
+`エラーを回収する`, `ツールを呼ぶ`, `したりできます`, and `設計の肝`.
+These additions use the exact surfaces listed in the YAML in both modes;
+they do not add general inflection coverage or ban everyday verbs. Existing
+`木`, `持つ`, `書く`, `別です`, and `ではありません` rules cover related examples.
+The shared criteria explain context-dependent terminology, precise operations,
+direct predicates, and objective tone. They distinguish invocation from
+execution, configuration text from configuration changes, assumptions from
+requirements, and collecting errors from catching them. Suggested alternatives
+are not automatic replacement rules. Start a new review session to include
+updated rules and criteria; existing sessions retain their saved versions.
+
+Version 0.8.4 also configures `呼ぶ`, `読み取る`, `拾う`, and the sahen predicate
+`回収する` as weak-predicate candidates. MeCab matches their inflections for
+any object, including `指示と宣言を読み取ります` and `最後のdetailsを拾います`.
+This broadens review cues, not mandatory terminology changes; everyday and
+technically valid uses can be retained with a specific reason. Literal mode
+matches only the selected surfaces in the YAML. Longer contextual phrases take
+precedence. `木全体` is an explicit contextual phrase in both modes because
+IPADIC can parse `木全` as a surname; this narrowly addresses that segmentation
+case without treating all names or compounds containing `木` as trees.
+
 Metaphorical-operation candidates include limited exact surfaces of
 `地味に効く`, `静かに壊れる`, `時間を溶かす`, and `側に倒す`, including selected polite,
 past, and connective forms listed in the packaged YAML. They use contextual
@@ -1057,12 +1096,12 @@ reports analyzer changes instead of treating results from different dictionaries
 as directly comparable. Prompt generation uses saved evidence and does not need
 MeCab. Changed rules or dictionary settings require a new scan.
 
-Sessions from earlier tool versions are not compatible with 0.8.2 (session schema 2). Keep the original baseline for
+Sessions from earlier tool versions are not compatible with 0.8.4 (session schema 2). Keep the original baseline for
 an ongoing review and finish it with the original version, or start a new review
 pass in a different directory:
 
 ```sh
-asciidoc-pubkit review scan book.adoc --output .pubkit/review-0.8.2
+asciidoc-pubkit review scan book.adoc --output .pubkit/review-0.8.4
 ```
 
 Severity describes review priority, not proof of an error. There is no AI-authorship

data/data/review-rules.ja.yml

@@ -59,6 +59,17 @@ terms:
     - 確かめます
     - 探します
     - 追えます
+    # Inspect retrieval, interpretation, and invocation for any object.
+    - 呼ぶ
+    - 呼びます
+    - 呼んで
+    - 呼んだ
+    - 読み取る
+    - 読み取ります
+    - 拾う
+    - 拾います
+    - 回収する
+    - 回収します
     # Construction, preparation, and combination.
     - 用意します
     - 作ります
@@ -135,6 +146,26 @@ terms:
     # Retain this specific phrase for existing allow lists and fuller evidence.
     - わけではありません
     - 別です
+    # Unusual technical translations; shared criteria own the alternatives.
+    - 木全体
+    - 構築入口
+    - 実装の読解
+    - エラーを回収する
+    - エラーを回収します
+    - 介入パターン
+    # Object-specific everyday verbs, not general bans on these verbs.
+    - ツールを呼ぶ
+    - ツールを呼びます
+    - 責務を持つ
+    - 指示を読み取る
+    - 指示を読み取ります
+    - 設定を書く
+    - 値を拾う
+    - 値を拾います
+    - 切り出す合図です
+    # Indirect capabilities and prerequisites; preserve their modality.
+    - したりできます
+    - することが前提です
     # Compressed noun relationships; suggestions are not automatic replacements.
     - 明示選択
     - 欠落理由
@@ -158,7 +189,8 @@ terms:
     - でもありません
     - してはいけません
     - 表してはいません
-    question: Check whether the phrase adds a condition that changes the reader's
+    question: Apply the shared prose criteria to terminology, operation wording,
+      indirect predicates, and qualifications. Check whether the phrase adds a condition that changes the reader's
       interpretation or action. If it only denies an overbroad reading already
       excluded by the main claim, revise that claim or remove the qualification.
       Preserve necessary negation, uncertainty, and technical boundaries. For
@@ -176,6 +208,10 @@ terms:
     - まとめると
     - このように
     - 要するに
+    # Author-centered discovery, indirect explanation, and unsupported emphasis.
+    - 見えてきます
+    - 理由がここにあります
+    - 設計の肝
     question: Check whether this opening or recap adds a claim, scope, or consequence
       beyond the surrounding prose. Keep useful framing; otherwise state the
       substantive claim directly. Do not force a uniform sentence pattern.
@@ -265,6 +301,12 @@ verbs:
   追う:
   - 追う
   - 追える
+  呼ぶ:
+  - 呼ぶ
+  読み取る:
+  - 読み取る
+  拾う:
+  - 拾う
   ある:
   - ある
 sahen:
@@ -278,6 +320,7 @@ sahen:
 - 区別
 - 解消
 - 混同
+- 回収
 negative_only:
 - 到達する
 - 意味する

data/data/writing/ja/criteria.md

@@ -66,6 +66,21 @@ Prefer a precise action or consequence:
 - `結果を残します` becomes `実行したコマンドと結果をプルリクエストへ記録します`
   when that is the storage location.
 
+For everyday verbs, identify the operation before choosing terminology:
+
+| Wording to inspect | Revision when supported by context |
+| --- | --- |
+| ツールを呼ぶ | 呼び出す for invocation; 実行する only for execution |
+| 責務を持つ | 責務を担う when describing responsibility, not stored state or ownership |
+| 指示を読み取る | 解析する for parsing or analysis; 抽出する for selecting information |
+| 設定を書く・設定を書きます | 設定を記述する・定義する for authoring a definition; 設定する for actually configuring something |
+| 値を拾う | 取得する for retrieval; 抽出する for selection from existing data |
+| 切り出す合図です | 切り出す目安です for a judgment criterion; シグナルです only for an actual signal |
+
+A familiar verb is not wrong merely because it is simple. Prefer the most direct
+accurate verb; do not introduce jargon or claim parsing, execution, or a state
+change that the evidence does not establish.
+
 Do not choose a more impressive synonym merely to vary sentence endings. Keep
 the technically correct verb when it names the real operation.
 
@@ -156,6 +171,27 @@ Common patterns that merit paragraph-level review include:
 Keep a contrast when the two alternatives and the difference that affects the
 decision are explicit. Keep necessary repetition when it improves technical
 accuracy.
+Inspect indirect predicates and author-centered emphasis as well:
+
+- Simplify `無効化したりできます` to `無効化できます` when it describes a single
+  capability. Preserve `たり` when it marks a meaningful non-exhaustive list.
+- For `提供することが前提です`, identify whose prerequisite or assumption is
+  stated. Use `提供する必要があります` only when the source establishes an actual
+  requirement; do not turn an assumption into an obligation.
+- For `宣言することは別です`, name what is compared and how it differs. A form such
+  as `宣言することとは異なります` needs an explicit, correctly oriented comparison.
+- For `そうではありません`, state the specific fact or negate the specific claim.
+  `実際には異なります` alone can retain the same missing referent. Preserve negation.
+- For `〜が見えてきます`, state the property the explanation makes identifiable.
+  `理解しやすくなります` is still a reader-effect claim that needs support; retain
+  literal visibility when discussing a display or visualization.
+- For `理由がここにあります`, connect the cause and result directly, for example
+  `これが〜の理由です` when `これ` has a clear referent.
+- For `設計の肝です`, identify the central responsibility or design decision.
+  `重要なポイントです` is not sufficient when it merely replaces one emphasis
+  phrase with another. Remove emotional, essay-like, or promotional emphasis
+  that supplies no technical fact; retain supported judgments and their basis.
+
 Do not shorten sentences or vary endings mechanically. Read the sequence aloud
 and repair the missing relationship rather than replacing one stock pattern
 with another.
@@ -170,12 +206,27 @@ glossary, and primary documentation before changing it.
 - Distinguish resource ownership from the responsibility to close or stop it.
 - Use the project's chosen forms consistently, such as `goroutine` versus
   `ゴルーチン`, or `reflection` versus `リフレクション`.
-- Replace unnecessary English words with established Japanese terms, but retain
-  identifiers and official names.
+- Prefer established terminology in the technical domain, whether Japanese or
+  katakana. Avoid forced literal translations and unnecessary English; retain
+  identifiers, official names, and the project's chosen terminology.
 - Do not describe a language operation with a loose everyday verb when the
   domain has a precise term, such as name shadowing, method promotion, parsing,
   serialization, cancellation, or resource release.
 
+Review unusual translations against the actual concept:
+
+| Wording to inspect | Context-dependent terminology |
+| --- | --- |
+| 木 | ツリー when referring to the data structure; keep literal trees and established mathematical usage |
+| 構築入口 | ビルドのエントリポイント only when the entry point starts a build, not object construction or initialization |
+| 実装の読解 | コードリーディング for reading code; 実装の解析 when analysis is the actual task |
+| エラーを回収する | エラーを捕捉する for catching; ハンドリングする for handling; preserve aggregation or collection when that is the behavior |
+| 介入パターン | フックパターン only when a hook mechanism is established, not arbitrary human intervention |
+
+These alternatives are not interchangeable replacements. Confirm the operation
+and the domain's established usage from code, the glossary, or primary
+documentation; record uncertainty rather than inventing a mechanism.
+
 Natural Japanese must not erase a technical distinction. If the original claim
 is ambiguous, inspect the code or authoritative documentation before editing.
 Preserve numerical values, versions, settings, responsibility, causal claims,
@@ -197,6 +248,15 @@ conditions, negation, numbers, actors, causes, and implementation requirements.
 For example, removing a unique constraint from a duplicate-message check can
 lose a concurrency requirement even when the revised paragraph sounds clearer.
 Record unresolved evidence and keep necessary repetition or qualifications.
+A review decision must come from reading the passage and checking the applicable
+criteria. A retained word does not establish a `keep` decision, and a removed
+word does not establish a satisfactory revision. Give passage-specific reasons
+for retaining terminology, conditions, and distinctions; leave unreviewed items
+pending rather than filling them with generic maintenance reasons. Compare the
+actual revision with the original claim, including paragraphs without machine
+candidates. Keep unresolved evidence and protected-scope proposals visible in
+the completion report.
+
 No candidate count or mechanical verification result proves semantic correctness.
 Sentence length, punctuation, and list density can help locate passages to read;
 they are not fixed acceptance thresholds. Keep lists for parallel items and

data/lib/asciidoc_pubkit/session.rb

@@ -184,6 +184,8 @@ module AsciidocPubkit
         After editing, reread each paragraph in context. Report unresolved issues and do not claim publication readiness.
         Mechanical verification does not establish semantic correctness.
 
+        #{review_process('paragraph', mode)}
+
         ## Shared prose criteria
 
         Apply these criteria only within the running-prose edit scope above. Their
@@ -318,6 +320,8 @@ module AsciidocPubkit
         original title, proposed title when applicable, and a short reason.
         Also track each candidate ID and its disposition. Do not claim publication readiness.
 
+        #{review_process('heading', mode)}
+
         ## Shared preservation and writing criteria
 
         The following common criteria provide meaning and terminology guidance.
@@ -372,6 +376,20 @@ module AsciidocPubkit
       text
     end
 
+    def review_process(unit, mode)
+      <<~TEXT
+        ## Review decisions and completion
+
+        Keep manifest.json, document.json, findings.json, and every baseline/ snapshot immutable. Restrict editing and formatting tools to the authorized source spans; never run them over the session directory.
+        Record each reviewed #{unit} ID and each candidate ID with revise, keep, or needs-evidence and a reason grounded in that passage. For keep, identify the concrete referent, operation, condition, necessary distinction, or established terminology that justifies retaining it. A shared generic reason does not establish individual review.
+        Never infer revise or keep from whether the matched string survives, disappears, or changes. Do not bulk-assign keep to unchanged #{unit}s or remaining candidates. An unreviewed item stays pending; use needs-evidence only after review identifies missing evidence.
+        Review #{unit}s without candidates against the shared criteria as well. For concerns outside the editable scope, report the location and a separate out-of-scope proposal; do not count protection as an editorial keep decision or edit protected content.
+        #{mode == 'revise' ? 'Compare the original and edited passages, check both information loss and unsupported additions, and reconcile each recorded decision with the actual diff. After verification, individually inspect newly detected or remaining candidates; retain the original candidate IDs in the review record.' : 'Compare each proposed revision with the original passage for information loss and unsupported additions. A revise decision means a proposed revision, not an applied edit.'}
+        Report mechanical preservation, candidate decisions, and #{unit}-level quality review separately. State selected and reviewed #{unit} counts, candidate decision counts, pending items, needs-evidence items, out-of-scope concerns, and evidence used. If any selected item or candidate remains pending, report the review as incomplete.
+        Mechanical verification does not read or validate these decision records. Its passed result and candidate counts do not establish review completion or writing quality; meaning_verified remains false. Do not claim semantic correctness from those checks.
+      TEXT
+    end
+
     def prompt_data(data, language)
       # Keep arbitrary manuscript text and custom settings inside their data fence.
       fence = '`' * [3, (data.scan(/`+/).map(&:length).max || 0) + 1].max

data/lib/asciidoc_pubkit.rb

@@ -9,7 +9,7 @@ require 'pathname'
 require 'optparse'
 
 module AsciidocPubkit
-  VERSION = '0.8.2'
+  VERSION = '0.8.4'
   class Error < StandardError; end
 
   def self.hash_text(text)

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: asciidoc-pubkit
 version: !ruby/object:Gem::Version
-  version: 0.8.2
+  version: 0.8.4
 platform: ruby
 authors:
 - CyberGarage