docs: generalize memory prompt record policy

docs/plan/memory-prompts.md

@@ -17,7 +17,7 @@ memory 関連 prompt は種別を問わず、最低限以下を共有する:
 - **単純 append を優先しない**。既存 record に統合できるなら update を優先する
 - **session 固有の進行状態を書かない**。長期参照価値のある内容だけを memory に残す
 - **既存 docs と重複保存しない**。`AGENTS.md`、`docs/plan/*`、固定運用文書に既にある内容を再保存しない
-- **git で追える事実を memory に書かない**。ticket file の作成・編集、TODO 更新、branch / worktree 操作、commit / merge / push、「commit X で実装した」「ticket Y を作った」「worker Pod を spawn した」等は git diff / log が真実で、memory に写すと陳腐化する。commit ハッシュ・branch 名・worktree パス・ticket file 名・PR 番号と組み合わせないと意味を成さない記録は採用しない
+- **特定の project-management 手法を前提にしない**。既定 prompt は issue tracker / task board / planning doc / changelog / version-control history / generated report などを「正確な内容の authoritative record」として一般化して扱う。memory はそれらの本文・status 変化・短命 identifier をそのまま複製する parallel ledger にはしない。一方で、今後の作業に役立つ持続的な project-management 事実、workflow 制約、優先度判断の根拠、複数 item にまたがる学びは採用してよい
 - **空出力を許容**する。保存価値が無ければ「何も追加しない」を正当な結果として扱う
 
 ### Phase 1: 活動抽出 prompt
@@ -32,10 +32,10 @@ Phase 1 は「派生物を作る」段階ではなく、「起きたことを抽
 
 ノイズ防御として、抽出時点で以下を除外する:
 
-- `attempts`: git で追える操作 (ticket file / TODO 編集、branch / worktree 作成、commit / merge / push、既知 ticket への worker Pod spawn) は除外。残すのは git からは復元できない情報 (ビルド/テスト結果、外部 API 応答、観測されたバグ再現、後段の判断材料となる設計実験結果) に限る
-- `discussions`: 当日中に陳腐化する一過性 triage (「次に着手するチケットはどれか」「いま review すべきか後でか」など) は除外。session を越えて意味を持つ論点 (アーキテクチャの trade-off、恒常的な制約、再来する問い) のみ残す
-- `decisions`: rationale が「この session で X をした」になるものは除外。設計 / 方針 / 取り組み方の根拠でない記録は decision ではなく作業ログ
-- 本文中に commit ハッシュ・branch 名・worktree パス・ticket file 名・PR 番号など陳腐化する identifier を埋め込まない
+- `attempts`: authoritative record の保守や外部 lifecycle の移動そのものだけを表す操作は除外。残すのはその record 自体からは復元できない情報 (ビルド/テスト結果、外部 API 応答、観測されたバグ再現、後段の判断材料となる設計実験結果、持続的な process lesson) に限る
+- `discussions`: 当日中に陳腐化する一過性 triage (直近の scheduling、status 確認、短命な順序決めなど) は除外。session を越えて意味を持つ論点 (アーキテクチャの trade-off、恒常的な workflow 制約、再来する問い) のみ残す
+- `decisions`: rationale が「この session で X をした」になるものは除外。設計 / 方針 / process / 取り組み方の根拠でない記録は decision ではなく作業ログ。ただし、作業管理上の持続的な方針や抽象化は decision として扱える
+- authoritative project record の title / body / checklist / raw status / 短命 identifier を本文に複製しない。正確な status mirror や identifier と組み合わせないと意味を成さない記録は採用しない
 
 ### Phase 2: 統合 + 整理 prompt
 

resources/prompts/internal/memory_consolidation_system.md

@@ -19,7 +19,7 @@ Your initial user message contains the staging entries, the full memory records,
 - **Update over create.** If an existing slug fits, edit it. Only create a new slug when no existing record fits and you can articulate why.
 - **`replaced` over delete.** When a Decision is superseded by a different one, mark the old one `status: replaced` with `replaced_by: <new-slug>`. Do not silently drop it.
 - **Don't duplicate static docs.** Skip content that already lives in `AGENTS.md`, `docs/plan/*`, or other fixed project documents.
-- **git is authoritative.** Do not record facts that git already tracks: ticket-file creation / edit, TODO updates, branch / worktree operations, commit / merge / push events, "implementation landed as commit X", "ticket Y was created", "worker Pod was spawned for Z". Diff and commit log are the truth there; memory shadowing it just rots. If a candidate write only makes sense when paired with a commit hash, branch name, worktree path, or ticket filename, drop it.
+- **Respect authoritative project records.** Issue trackers, task boards, planning documents, changelogs, version-control history, and generated reports are authoritative for their exact contents. Do not mirror them verbatim, preserve raw status churn, or maintain a parallel ledger in memory. Durable project-management facts and abstractions are valid when they help future work: recurring workflow constraints, prioritisation rationale, long-lived ownership / process decisions, and stable lessons that cut across individual items. If a candidate write only makes sense as an exact status mirror or when paired with a transient identifier, drop it.
 - **Empty output is fine.** If a staging entry doesn't justify a memory write, skip it.
 - **Slug rules.** Slugs are kebab-case, short, recognisable, and must be unique within their kind. Same-slug create is a linter error — use Edit instead.
 - **Linter errors come back as tool errors.** When the memory linter rejects a write, read the error, fix the issue (missing frontmatter field, oversized body, unknown reference, etc.), and try again. Do not work around the rule.

resources/prompts/internal/memory_extract_system.md

@@ -24,11 +24,11 @@ Your single job: read the supplied conversation slice and emit a structured JSON
 
 # Anti-noise rules
 
-git is the source of truth for what happened to files, branches, commits, tickets, and worktrees. Memory must NOT shadow it.
+Authoritative project records (issue trackers, task boards, planning documents, changelogs, version-control history, generated reports) are the source of truth for their exact contents. Memory must not mirror those records verbatim or maintain a parallel status ledger, but it may capture durable project-management facts, workflow constraints, recurring patterns, and abstractions when they will help future work.
 
-- `attempts`: skip any action whose substance is a git-trackable operation — creating / editing a ticket file, adding a TODO entry, opening a branch / worktree, running `commit` / `merge` / `push`, spawning a worker Pod for a known ticket. The corresponding diff / commit log already records it. Keep `attempts` for things that are NOT recoverable from git: build / test outcomes, external API responses, observed bug reproductions, design experiments whose results inform later judgement.
-- `discussions`: skip transient triage that goes stale within the day — "which ticket to start next", "should we review now or later", checklist-style status reads. Keep discussions whose points outlive the session (architectural trade-offs, durable constraints, recurring questions).
-- `decisions`: the rationale must be a design / policy / approach reason, not "we did X in this session". Recording "a ticket was created for Y" or "implementation landed as commit Z" is NOT a decision — those belong to git, not memory.
-- Do not embed identifiers that age out of relevance: commit hashes, branch names, worktree paths, ticket file names, PR numbers. If a record is only meaningful with such an identifier, the record itself is probably session-local and should be skipped.
+- `attempts`: skip actions whose only substance is maintaining an authoritative record or moving an item through an external lifecycle. Keep attempts for outcomes that are not captured by that record itself: build / test outcomes, external API responses, observed bug reproductions, design experiments, and process lessons that inform later judgement.
+- `discussions`: skip transient triage that goes stale within the day — immediate scheduling, checklist-style status reads, or short-lived sequencing choices. Keep discussions whose points outlive the session (architectural trade-offs, durable process constraints, recurring workflow questions).
+- `decisions`: the rationale must be a design / policy / process / approach reason, not "we did X in this session". Recording that an item was filed, completed, or moved through a lifecycle is NOT a decision; recording the durable policy or abstraction behind that workflow can be.
+- Avoid copying titles, bodies, checklists, raw statuses, or short-lived identifiers from authoritative project records. If a record is only meaningful as an exact status mirror or with a transient identifier, the record itself is probably session-local and should be skipped.
 
 When you have produced the JSON, call `write_extracted` and end the turn. No follow-up text.

tickets/memory-prompts-remove-ticket-policy.md

@@ -44,3 +44,8 @@
 - `resources/prompts/internal/memory_consolidation_system.md`
 - `docs/plan/memory-prompts.md`
 - `docs/plan/memory.md`
+
+## Review
+- 状態: Approve
+- レビュー詳細: [./memory-prompts-remove-ticket-policy.review.md](./memory-prompts-remove-ticket-policy.review.md)
+- 日付: 2026-05-09

tickets/memory-prompts-remove-ticket-policy.review.md

@@ -0,0 +1,33 @@
+# Review: Memory prompts: project-specific ticket shadow policy を外す
+
+## 前提・要件の確認
+
+- bundled memory prompts から ticket / TODO 固有の shadow 禁止を削除: OK。Phase 1 prompt は ticket / TODO / worktree / branch / commit 等の列挙を消し、authoritative project records 一般へ置き換えている（`.worktree/memory-prompts-remove-ticket-policy/resources/prompts/internal/memory_extract_system.md:27-32`）。Phase 2 prompt も `git is authoritative` と ticket/TODO 固有列挙を削除し、issue trackers / task boards / planning documents 等の一般表現へ置換している（`.worktree/memory-prompts-remove-ticket-policy/resources/prompts/internal/memory_consolidation_system.md:21-22`）。
+- `git is authoritative` の扱いを一般ユーザー向けに弱める: OK。正確な内容の authoritative record を memory が mirror しないという原則は残しつつ、durable project-management facts / workflow constraints / prioritisation rationale / process decisions / cross-item lessons は保存可能と明示している（`memory_extract_system.md:27-32`, `memory_consolidation_system.md:22`）。
+- Phase 1 / Phase 2 の整合: OK。Phase 1 の抽出除外条件と Phase 2 の統合時ルールがどちらも「raw status ledger は作らないが、持続的な抽象は採用可」という同じ方針になっている（`memory_extract_system.md:29-32`, `memory_consolidation_system.md:22`）。
+- default prompt が特定 project-management 手法を前提にしないことの文書化: OK。`docs/plan/memory-prompts.md` の共通原則に「特定の project-management 手法を前提にしない」が追加され、ticket / TODO 固有の旧記述が一般化されている（`.worktree/memory-prompts-remove-ticket-policy/docs/plan/memory-prompts.md:19-20`）。
+- 既存テスト / prompt catalog 検査: OK。`cargo test -p pod prompt::catalog` を実行し、11 tests passed。
+
+## アーキテクチャ・スコープ
+
+- 変更は bundled prompt 2 件と plan doc 1 件のみで、memory の file format / linter / Phase 1 / Phase 2 実行機構には触れていない。ticket の範囲外を守っている。
+- prompt 文言の責務分離は妥当。INSOMNIA 固有の ticket shadow 回避を default prompt から外し、workspace / user prompt override 側へ逃がせる状態にしている。
+- `docs/plan/memory-prompts.md` も同時に更新しており、実装 prompt と計画文書が乖離していない。
+
+## 指摘事項
+
+### Blocking
+
+なし。
+
+### Non-blocking / Follow-up
+
+なし。
+
+### Nits
+
+- `docs/plan/memory-prompts.md:68` に「git で可逆」という表現が残っているが、これは tidy phase の drop 判断に関する一般的な VCS 可逆性の話で、ticket / TODO 固有の shadow policy ではないため本チケットの完了を妨げない。
+
+## 判断
+
+Approve — ticket / TODO 固有の bundled prompt 方針は削除され、一般的な authoritative record 非重複方針へ置き換わっている。変更範囲も prompt / plan に限定され、既存 prompt catalog tests も通っている。