architext 0.0.3 → 0.0.5

package/dist/templates/en/docs/prompts/ref.md

@@ -0,0 +1,163 @@
+<protocol_ref>
+**Trigger**: `/archi.ref <sub> [args]`
+**Goal**: Manage external knowledge refs (3rd-party APIs, internal SDKs, business rules, etc.); store in `refs/` in structured form; inject into `plan`/`code` on demand.
+
+**Subcommands**:
+| Subcommand | Format | Purpose |
+|:---|:---|:---|
+| `add` | `/archi.ref add [input]` | Summarize and store external knowledge (core) |
+| `list` | `/archi.ref list` | List all stored refs |
+| `update` | `/archi.ref update <id>` | Re-summarize specified ref |
+| `remove` | `/archi.ref remove <id>` | Remove specified ref |
+
+<meta>
+  <style>Analytical, Precise, Context-Aware</style>
+  <language>English</language>
+  <principles>
+    1. **Summarize, Not Copy**: Do not copy full original; must distill into structured summary (key interfaces / constraints / examples) for efficient AI use; compression ratio ≥ 50%.
+    2. **Format-Aware**: Auto-select best carrier format (`.md` / `.json` / `.yaml`) by content type; preserve original format semantics.
+    3. **Tag-Driven**: Each ref must carry tags for `plan`/`code` to match and inject; no tagless refs.
+    4. **Index-First**: All ops must sync `refs/index.json`; AI loads via index on demand; no full scan.
+  </principles>
+</meta>
+
+<!-- ─── Format selection rules ─── -->
+<format_selection>
+  **Format-Aware rules** (referenced during step_3_store):
+
+  | Content type | Recommended format | Rationale |
+  |:---|:---|:---|
+  | Web / URL / PDF / plain text | `.md` | Structured summary; highest AI read efficiency |
+  | OpenAPI / Swagger spec | `.yaml` | Preserve machine-readable structure; do not convert to .md |
+  | JSON Schema / config files | `.json` | Original structured data format optimal |
+  | Mixed (many code examples) | `.md` (with code blocks) | Code blocks keep syntax; add explanatory context |
+  | User paste (pure Markdown) | `.md` | Same format store; may refine |
+</format_selection>
+
+<!-- ═══════════════════════════════════════════════ -->
+<!--                   ADD subcommand               -->
+<!-- ═══════════════════════════════════════════════ -->
+
+<sub_add>
+
+<step_0_ingest>
+  **Role**: Intelligence Analyst
+  **Trigger**: `/archi.ref add [input]`
+  **Action**: Parse `[input]` to determine input source.
+
+  | input form | Handling |
+  |:---|:---|
+  | Local file path (e.g. `./docs/api.yaml`) | Read file content; record `sourceType: local-file` |
+  | URL (e.g. `https://...`) | Fetch page content; record `sourceType: url` |
+  | Not provided (conversation mode) | Ask user to paste content or provide path/URL; wait for input; record `sourceType: manual` |
+
+  Check if `[[__DOCS_DIR__]]/refs/index.json` exists:
+  - Exists → read; get existing id list (avoid duplicate names)
+  - Not exists → initialize as `{ "refs": [] }`
+
+  **Output**: Internal (raw content + sourceType + existing id list); proceed to `<step_1_analyze>`.
+</step_0_ingest>
+
+<step_1_analyze>
+  **Action**:
+  1. **Content type identification**: Determine raw content type (per `<format_selection>`); decide recommended storage format.
+  2. **Key info extraction**: Core interfaces/endpoints/signatures, params and types, important constraints/limits, auth (if any), typical examples (≤3).
+  3. **Info gap identification**: id naming (infer candidate), tags (from standard: `api`/`sdk`/`internal`/`payment`/`auth`/`map`/`notification`/`storage`/custom), focus areas (when content large, confirm user's priority interfaces).
+
+  **Output**: Internal analysis summary; proceed to `<step_2_interview>` (when gaps) or directly `<step_3_store>` (when complete).
+</step_1_analyze>
+
+<step_2_interview>
+  **Trigger**: Only when id / tags / focus uncertain.
+  **Action**: Ask user; max 3 questions; options preferred.
+
+  **Output**: Ref info confirmation — include content type and storage format, content summary, Q1 ref ID (AI inferred candidate + custom), Q2 tags (multi-select), (when content large) Q3 focus areas.
+
+  **Gate**: Wait for user reply before `<step_3_store>`.
+</step_2_interview>
+
+<step_3_store>
+  **Action**:
+  1. Determine storage params: `id` (user confirmed or AI inferred), `format` (from format_selection), `filename: {id}.{format}`, `tags`.
+
+  2. **Generate ref file content** (by format):
+     - **`.md`**: frontmatter (id/title/tags/sourceType/source/created/updated) + 4 sections (core info, key interfaces table, important constraints, examples ≤3)
+     - **`.yaml`**: Store trimmed OpenAPI/Swagger (remove redundant example fields; keep schema)
+     - **`.json`**: Store original structure (remove comments; keep structure)
+
+  3. **Write file**: `[[__DOCS_DIR__]]/refs/{id}.{ext}`
+
+  4. **Update index**: Append to `[[__DOCS_DIR__]]/refs/index.json` refs array (id/title/tags/format/file/sourceType/updatedAt).
+
+  **Output**:
+  ```
+  ADDED  [[__DOCS_DIR__]]/refs/{id}.{ext}
+  MODIFIED  [[__DOCS_DIR__]]/refs/index.json
+  ```
+</step_3_store>
+
+<step_4_signoff_add>
+  **Output**: Add summary — ref ID, storage format and rationale, tags, file path, usage notes (plan: auto-read / code: supplement context / manual ref).
+</step_4_signoff_add>
+
+</sub_add>
+
+<!-- ═══════════════════════════════════════════════ -->
+<!--                  LIST subcommand               -->
+<!-- ═══════════════════════════════════════════════ -->
+
+<sub_list>
+
+**Trigger**: `/archi.ref list`
+**Action**: Read `[[__DOCS_DIR__]]/refs/index.json`.
+
+| Condition | Handling |
+|:---|:---|
+| Index missing / refs empty | Prompt "No refs yet; run `/archi.ref add` to add first" |
+| Normal | Display grouped by tags (ID / title / format / updated) |
+
+</sub_list>
+
+<!-- ═══════════════════════════════════════════════ -->
+<!--                 UPDATE subcommand              -->
+<!-- ═══════════════════════════════════════════════ -->
+
+<sub_update>
+
+**Trigger**: `/archi.ref update <id>`
+**Action**:
+1. Find `<id>` in index.json; get `file` and `sourceType`.
+2. If `sourceType` is `url` → re-fetch; `local-file` → re-read; `manual` → prompt paste new content.
+3. Re-run `<step_1_analyze>` + `<step_3_store>` (keep original id/tags/format; refresh content and `updatedAt` only).
+
+| Condition | Handling |
+|:---|:---|
+| id not in index.json | Stop — prompt to check id; run `/archi.ref list` to view |
+
+</sub_update>
+
+<!-- ═══════════════════════════════════════════════ -->
+<!--                 REMOVE subcommand              -->
+<!-- ═══════════════════════════════════════════════ -->
+
+<sub_remove>
+
+**Trigger**: `/archi.ref remove <id>`
+**Action**:
+1. Find and remove `<id>` entry from index.json.
+2. Delete corresponding `refs/{id}.{ext}` file.
+3. Update index.json.
+
+**Output**:
+```
+REMOVED  [[__DOCS_DIR__]]/refs/{id}.{ext}
+MODIFIED  [[__DOCS_DIR__]]/refs/index.json
+```
+
+| Condition | Handling |
+|:---|:---|
+| id not found | Stop — prompt to check id; run `/archi.ref list` to view |
+
+</sub_remove>
+
+</protocol_ref>

package/dist/templates/en/docs/prompts/remove.md

@@ -1,161 +1,109 @@
 <protocol_remove>
   **Trigger**: `/archi.remove <id>`
-  **Goal**: Fully decommission the specified Task — delete docs, code, and global references with zero residue.
+  **Goal**: Decommission the specified Task from the project — delete docs, code, global refs; ensure zero residue.
 
 <meta>
     <style>Surgical, Cautious, Thorough</style>
     <language>English</language>
     <principles>
-      1.  **Dependency Safety**: If other tasks depend on this Task, dependencies must be resolved before proceeding.
-      2.  **Confirmation Gate**: Deletion is irreversible; must present full impact and obtain explicit confirmation.
-      3.  **Zero Residue**: Cleanup must cover all associated layers (docs/code/global data/glue entries).
-      4.  **No Side Effects**: Only remove content related to the target Task; forbidden to touch unrelated files.
+      1.  **Dependency Safety**: When other tasks depend on this Task, must decouple first to continue.
+      2.  **Confirmation Gate**: Delete is irreversible; must show full impact and get explicit user confirmation.
+      3.  **Zero Residue**: Cleanup must cover all layers (docs/code/global data/addressing entries).
+      4.  **No Side Effects**: Only delete target Task content; do not touch unrelated files.
     </principles>
 </meta>
 
 <step_1_resolve>
-    **Role**: System Analyst
     **Action**:
-    1.  **Resolve ID**: Parse `<id>` from `[[__DOCS_DIR__]]/global/roadmap.json` → Task Name, Slug, status.
-    2.  **ID Not Found** → Error with list of available task IDs.
-    3.  **Load Context**:
-        - `[[__DOCS_DIR__]]/tasks/<id>_<Slug>/` — All docs (spec.md, ui.md, plan.json, etc.)
-        - `[[__DOCS_DIR__]]/global/roadmap.json` — Task dependency graph
-        - `[[__DOCS_DIR__]]/global/map.json` — Architecture registry
-        - `99_context_glue.md` — Code↔doc associations
+    1.  **Resolve ID**: Parse `<id>` from roadmap.json → Task Name, Slug, status.
+    2.  **ID Not Found** → Error and list available task IDs.
+    3.  **Load**: task docs directory, roadmap.json (deps), map.json (architecture reg), 99_context_glue.md (associations).
 
     **Output**: Target Task info (ID, name, status, associated file count).
 </step_1_resolve>
 
 <step_2_impact>
-    **Role**: Impact Assessor
-    **Action**: Scan each layer to produce an impact manifest.
+    **Action**: Scan layer by layer; generate impact list.
 
-    ### 2.1 Dependency Check (Blocking)
+    ### 2.1 Dependency check (blocking)
 
-    Scan all tasks' `deps` in `roadmap.json` for references to `<id>`.
+    Scan roadmap.json all tasks `deps`; find tasks depending on `<id>`.
 
-    | Situation | Handling |
+    | Condition | Handling |
     |:---|:---|
-    | No downstream deps | Pass, continue |
-    | Has downstream deps with `pending`/`blocked` status | List affected tasks; ask user: remove dependency and continue, or abort |
-    | Has downstream deps with `active`/`done` status | **Block** — these tasks may use this Task's interfaces/components. Must `/archi.edit` to decouple first |
+    | No downstream deps | Pass |
+    | Has downstream deps and `pending`/`blocked` | List; ask: remove dep and continue or abort |
+    | Has downstream deps and `active`/`done` | **Block** — must `/archi.edit` to decouple first |
 
-    ### 2.2 Code File Identification
+    ### 2.2 Code file identification
 
-    Locate code files belonging to this Task via:
-    1.  `99_context_glue.md` code paths associated with `<id>`.
-    2.  `map.json` modules/directories registered for this Task.
-    3.  `plan.json` file paths mentioned in tasks.
-    4.  Scan code directories for files named with Slug or clearly scoped to this Task.
+    Locate code files via: context_glue paths, map.json registered modules, plan.json mentioned files, Slug-named or clearly owned files.
 
-    ### 2.3 Global Reference Scan
+    ### 2.3 Global ref scan
 
-    | File | Scan Target |
+    | File | Scan content |
     |:---|:---|
-    | `roadmap.json` | Task entry + `deps` references in other tasks |
-    | `map.json` | Module registry entries + `featureRelations` entries where the deleted Task is the aggregator |
+    | `roadmap.json` | Task entry + deps refs |
+    | `map.json` | Module entry + featureRelations where deleted Task is aggregator |
     | `99_context_glue.md` | Association entries |
-    | `dictionary.json` | Terms exclusive to this Task (flag only, no auto-delete) |
-    | `error_codes.json` | Error codes exclusive to this Task (flag only, no auto-delete) |
+    | `dictionary.json` | Exclusive terms (mark only) |
+    | `error_codes.json` | Exclusive error codes (mark only) |
 
-    ### 2.4 Aggregation Linkage Check
+    ### 2.4 Aggregator linkage check
 
-    Read `map.json.featureRelations`; determine whether the deleted Task falls within any aggregator's `sources` coverage.
+    [[SUBAGENT: archi-feature-relations|mode: cleanup, context: Check deleted Task refs in featureRelations; output impact report]][[NO-SKILL: (Skill not installed: read `[[__DOCS_DIR__]]/skills/archi-feature-relations/SKILL.md`, follow mode: cleanup logic)]]
 
-    | Situation | Handling |
-    |:---|:---|
-    | Not in any aggregator's sources range | No special handling |
-    | In an aggregator's sources range | List in impact report; prompt to check if aggregator content needs to be cleaned up after deletion |
-    | Deleted Task is itself an aggregator | Also remove its entry from `featureRelations` |
-
-    ### 2.5 Cross-Task References
-
-    Scan other Tasks' `spec.md` for references to the target Task's interfaces, components, or data. Flag as `[Breaking]`.
-
-    **Output**: Impact report to user:
-    ```
-    ### Decommission Impact Report: <ID> <Name>
-
-    **Task Status**: [status]
-
-    **Docs to delete**:
-    - [[__DOCS_DIR__]]/tasks/<id>_<slug>/  (N files)
-
-    **Code to delete**:
-    | File/Directory | Source |
-    |:---|:---|
-    | src/xxx/xxx.ts | context_glue |
-    | src/xxx/ | map.json |
+    ### 2.5 Cross-Task refs
 
-    **Global reference cleanup**:
-    - roadmap.json: remove task <id>, clean N deps references
-    - map.json: remove N module entries
-    - context_glue: remove N associations
+    Scan other Tasks' `spec.md`; check refs to deleted Task's interface/components/data. Mark refs as `[Breaking]`.
 
-    **[?present] Terms/error codes residue** (manual confirmation needed):
-    - dictionary.json: [term1], [term2]
-    - error_codes.json: [ERR_XXX]
+    **Output**: Output decommission impact report to user — include Task status, docs and code to delete (file/source table), global ref cleanup items (roadmap/map/context_glue), (if any) term/error code residue (manual confirm needed), (if any) aggregator linkage table, (if any) cross-Task ref [Breaking] table. End: OK to execute / Abort to cancel.
 
-    **[?present] Aggregation linkage** (check needed):
-    | Aggregator | checkNote |
-    |:---|:---|
-    | [aggregator ID/path] | [checkNote content] |
-
-    **[?present] Cross-Task references [Breaking]**:
-    | Referencing Task | Reference Content | Suggestion |
-    |:---|:---|:---|
-    | FEAT-005 spec.md | Calls <id>'s UserAPI | /archi.edit FEAT-005 to decouple first |
-
-    ---
-    > Reply **OK** to confirm; reply **abort** to cancel.
-    ```
-
-    **Gate**: Proceed to step_3 after user replies OK. Re-warn if `[Breaking]` cross-Task refs exist.
+    **Gate**: Proceed to step_3 only after user replies OK. Re-warn when `[Breaking]` refs exist.
 </step_2_impact>
 
 <step_3_execute>
     **Role**: Surgeon
-    **Action**: Execute in this order (order is mandatory).
+    **Action**: Execute in this order (order immutable).
 
-    | # | Operation | Target |
+    | # | Action | Target |
     |:---|:---|:---|
-    | 1 | Delete code files/directories | Code paths from step_2 |
-    | 2 | Delete Task doc directory | `[[__DOCS_DIR__]]/tasks/<id>_<slug>/` |
-    | 3 | Update `roadmap.json` | Remove task entry; clean `deps` refs to `<id>` in other tasks |
-    | 4 | Update `map.json` | Remove module entries; if this Task is an aggregator, also remove its `featureRelations` entry |
-    | 5 | Update `99_context_glue.md` | Remove association entries for this Task |
+    | 1 | Delete code files/dirs | Code paths from step_2 |
+    | 2 | Delete Task docs dir | `[[__DOCS_DIR__]]/tasks/<id>_<slug>/` |
+    | 3 | Update `roadmap.json` | Remove task entry; clean deps refs |
+    | 4 | Update `map.json` | Remove module entry + featureRelations entry |
+    | 5 | Update `99_context_glue.md` | Remove association entries |
     | 6 | [?exclusive terms] Update `dictionary.json` | Remove or mark deprecated |
-    | 7 | [?exclusive codes] Update `error_codes.json` | Remove or mark deprecated |
-    | 8 | [?aggregation linkage] Check aggregator code | Verify aggregator no longer references the deleted Task |
+    | 7 | [?exclusive error codes] Update `error_codes.json` | Remove or mark deprecated |
+    | 8 | [?aggregator linkage] Check aggregator code | Confirm refs cleaned |
 
-    Log each operation (file path + operation type) as it completes.
+    Record operation log after each step.
 </step_3_execute>
 
 <step_4_verify>
-    **Role**: Verification Engineer
-    **Terminal Gate** (Do not skip; must complete before step_5 output):
+    **Terminal Gate** (do not skip): Standard check (task --check + render).
     | Step | Command | Pass Condition |
     |:---|:---|:---|
-    | 1 | `npx archi task --check` | No ERROR-level issues, no dangling deps |
-    | 2 | `npx archi render` | `.md` views generated |
-    | 3 | Run project build command | Zero compilation errors |
-
-    | Check | Pass Criteria |
-    |:---|:---|
-    | Roadmap consistency | `--check` passes, no dangling deps |
-    | Build | Zero compilation errors |
-    | Residual refs | No import/require of deleted modules in codebase |
+    | 3 | Run project build command | Zero compile errors; no import/require to deleted module |
 
-    Build failure or residual refs found → locate, fix, and recheck.
+    If build fails or residual refs found → locate, fix, re-check.
 </step_4_verify>
 
 <step_5_summary>
-    **Output**: Decommission summary:
+    **Pre-signoff Checklist** (confirm each item before Output):
+    □ Code files/directories — deleted (step_3 action #1)
+    □ tasks/<id>_<slug>/ doc directory — deleted (step_3 action #2)
+    □ roadmap.json — task entry + deps refs — cleaned (step_3 action #3)
+    □ map.json — module entry + featureRelations — cleaned (step_3 action #4)
+    □ 99_context_glue.md — association entries — cleaned (step_3 action #5)
+    □ (if exclusive terms/codes) dictionary.json + error_codes.json — handled
+    □ Terminal Gate — project build passes, no residual import refs (step_4)
+
+    **Output**: Decommission completion summary:
     - **Deleted**: N doc files, N code files
-    - **Cleaned**: roadmap / map / context_glue references
-    - **Build status**: pass/fail
-    - **[?present] Manual follow-up needed**: terms/error codes/cross-Task ref residue
+    - **Cleaned**: refs in roadmap / map / context_glue
+    - **Build status**: Pass/Fail
+    - **[?if any] Manual follow-up**: Term/error code/cross-Task ref residue
     - **Git Commit Suggestion**: `feat(remove): decommission <ID> <Name>`
 </step_5_summary>
 

package/dist/templates/en/docs/prompts/revise.md

@@ -1,160 +1,117 @@
 <protocol_revise>
   **Trigger**: `/archi.revise [context]`
-  **Goal**: Execute project-wide global changes (tech stack, architecture, vision, etc.), analyze impact scope, modify global assets after user confirmation, and cascade updates to affected Task documents.
+  **Goal**: Execute project-wide global changes (tech stack, architecture, vision, etc.); analyze impact scope; modify global assets after user confirmation and cascade update affected Task docs.
 
 <constraints_cursor>
-    **Mode Lock**: This protocol must run in **Agent Mode (Normal Mode)**. Prohibited from switching to Plan Mode or other read-only modes.
+    **Mode Lock**: This protocol must run in **Agent Mode (Normal Mode)**. Do not switch to Plan Mode or other read-only modes.
 </constraints_cursor>
 
 <meta>
     <style>Strategic, Analytical, Cautious, Traceable</style>
     <language>English</language>
     <principles>
-      1.  **User Gate**: Global changes require item-by-item user confirmation before execution. Prohibited from modifying without approval.
-      2.  **Impact First**: Output complete impact analysis before making any changes.
-      3.  **Doc Cascade**: After global asset changes, must update affected Task docs following edit protocol standards.
+      1.  **User Gate**: Global changes require item-by-item user confirmation before execution. Do not modify without approval.
+      2.  **Impact First**: Output complete impact analysis before making changes.
+      3.  **Doc Cascade**: After global asset changes, must sync affected Task docs per edit standards.
       4.  **Traceability**: Every change must be traceable — state rationale and impact scope.
     </principles>
 </meta>
 
 <step_1_load>
-    **Role**: Systems Analyst
     **Action**:
-    1.  **Read Global Assets**:
-        - `[[__DOCS_DIR__]]/global/vision.md`
-        - `[[__DOCS_DIR__]]/global/roadmap.json`
-        - `[[__DOCS_DIR__]]/global/map.json`
-        - `[[__DOCS_DIR__]]/global/dictionary.json`
-        - `[[__DOCS_DIR__]]/global/error_codes.json`
-        - `02_tech_stack.md`
-        - [?UI] `[[__DOCS_DIR__]]/global/design_tokens.json`
-        - [?Data] `[[__DOCS_DIR__]]/global/data_snapshot.json`
-    2.  **Scan Task Index**: Scan `[[__DOCS_DIR__]]/tasks/` directory, build Task index (ID, name, status).
-    3.  **Intent Analysis**: Based on user `[context]`, identify which global asset categories are affected.
-
-    **Output**: Change intent summary — list of initially identified global files. Proceed to step_2.
+    1.  **Load**: All global assets (vision/roadmap/map/dictionary/error_codes/02_tech_stack + conditional: design_tokens/data_snapshot).
+    2.  **Scan Task Index**: Scan tasks/ directory; build Task index (ID, name, status).
+    3.  **Intent Analysis**: From `[context]`, identify affected global asset categories.
+
+    **Output**: Change intent summary — list initially identified global files; proceed to step_2.
 </step_1_load>
 
 <step_2_interview>
     **Role**: Chief Architect
     **Action**:
-    Clarify change boundaries based on user description and loaded context:
+    Clarify change boundaries from user description and loaded context:
 
     | Dimension | Description |
     |:---|:---|
-    | Change Scope | Precisely identify which global files and rules are affected |
-    | Change Motivation | Why this change is needed, what outcome is expected |
-    | Exclusion List | Any Tasks that should not be affected |
+    | Change scope | Precisely identify which global files and rules are affected |
+    | Change motivation | Why this change; expected outcome |
+    | Exclusion list | Any Tasks that should not be affected |
 
-    - Context is clear enough → merge into step_3 and output impact analysis directly.
-    - Ambiguity or multiple viable directions → ask user to confirm (A/B/C/D options).
+    - When context is clear enough → merge into step_3 and output impact analysis directly.
+    - When ambiguity or multiple viable directions → ask user to confirm (A/B/C/D options).
 </step_2_interview>
 
 <step_3_impact>
-    **Role**: Impact Analyst
-    **Constraint**: **Output only, prohibited from modifying any files**. This step produces the "Change Impact Assessment".
-
-    **Output Format**:
-    ```
-    ## Change Impact Assessment: [Change Topic]
-
-    ### 1. Global Asset Change List
-    | File | Change Content | Type |
-    |:---|:---|:---|
-    | vision.md | [what specifically changes, or "no change needed"] | modify/no change |
-    | 02_tech_stack.md | [...] | ... |
-    | roadmap.json | [...] | ... |
-    | map.json | [...] | ... |
-    | dictionary.json | [...] | ... |
-    | data_snapshot.json | [...] | ... |
-    | design_tokens.json | [...] | ... |
-    | error_codes.json | [...] | ... |
-
-    ### 2. Affected Tasks List
-    | Task ID | Name | Impact Points | Severity |
-    |:---|:---|:---|:---|
-    | INF-001 | [name] | [which parts of spec/ui/plan are affected] | High/Med/Low |
-
-    ### 3. Items Requiring User Decision (if any)
-    - [?] [uncertain item description + options]
-    - [?] ...
-
-    ---
-    > Reply **OK** to confirm all; or annotate items that need adjustment.
-    ```
-
-    **Gate**: Wait for user confirmation. Prohibited from entering step_4 without confirmation.
+    **Constraint**: **Output only; do not modify any files**. This step produces the "Change Impact Assessment".
+
+    **Output**: Change impact assessment — global asset change list table (file/change content/type), affected Task list table (ID/name/impact points/severity), user decision items (if any). End with confirmation: OK to confirm all; or annotate adjustments.
+
+    **Gate**: Wait for user confirmation. Do not enter step_4 without confirmation.
 </step_3_impact>
 
 <step_3_5_refinement>
-    **Role**: Consultant
-    **Trigger**: User replies non-OK, contains corrections, rejections, or additional input.
-    **Action**: Do not execute changes. Incorporate user feedback, refresh and re-output the impact assessment. Wait for re-confirmation.
+    **Trigger**: User reply is not OK — contains corrections, rejections, or additional input.
+    **Action**: Do not execute changes. Incorporate feedback, refresh impact assessment and re-output; await re-confirmation.
 </step_3_5_refinement>
 
 <step_4_execute>
-    **Role**: Execution Engineer
     **Action**:
 
-    **Safety Checkpoint** (required before execution):
+    **Safety Checkpoint** (must complete before execution):
     1. Check Git working directory status (run `git status`).
-    2. If uncommitted changes exist → prompt user to commit or stash first, then continue.
-    3. Once working directory is clean, inform user: to roll back, run `git checkout -- .` to restore pre-change state.
+    2. If uncommitted changes exist → prompt user to commit or stash first.
+    3. Once working directory clean, inform user: to rollback, run `git checkout -- .`.
 
-    **Phase 1 — Modify Global Assets**:
-    Modify global files per the user-confirmed list. Output change summary for each file.
+    **Phase 1 — Modify global assets**:
+    Modify global files per user-confirmed list. Output change summary for each file.
 
-    **[?UI] Phase 1.5 — Design System Change Check**:
-    If `design_tokens.json` has any of the following changes, notify the user after Phase 2:
+    **(UI projects only) Phase 1.5 — Design system change check**:
+    If `design_tokens.json` has the following changes, notify user to re-run `archi-ui-wireframe` to regenerate:
 
-    | Change scope | Impact | Action |
-    |:---|:---|:---|
-    | `primitivePalette.brand` / `semanticTokens.colors` | Brand/semantic color change | Notify user: re-run `archi-ui-wireframe` Phase 2 (full re-coloring) |
-    | `semanticTokens.typography` | Font change | Notify user: re-run Phase 2 |
-    | `motion.preference` / `motion.patterns` | Motion change | Notify user: re-run Phase 2 |
-    | `illustration.iconLibrary` | Icon library change | Notify user: re-run Phase 2 |
-    | `layout` (radius/spacing/shadow) | Component size/radius change | Notify user: re-run Phase 2 |
+    | Change scope | Impact |
+    |:---|:---|
+    | `primitivePalette.brand` / `semanticTokens.colors` | Brand/semantic color change |
+    | `semanticTokens.typography` | Font change |
+    | `motion.preference` / `motion.patterns` | Motion change |
+    | `illustration.iconLibrary` | Icon library change |
+    | `layout` (radius/spacing/shadow) | Component size/radius change |
 
-    > If none of the above fields changed (e.g., only `mode.default` changed), Phase 2 re-run is not needed.
+    > If above fields unchanged (e.g. only `mode.default` changed), no need to re-run.
 
-    **Phase 2 — Cascade Update Task Docs**:
+    **Phase 2 — Cascade update Task docs**:
     For each affected Task, follow `/archi.edit` standards:
-    1.  Update `spec.md` (logic/rules that need adjustment due to global changes).
-    2.  [?UI] Update `ui.md` (scope/interaction adjustments due to global changes); if screen structure is affected, run `archi-ui-wireframe` Skill (incremental update mode) to sync both `ui_concept.html` + `ui_context.md`.
-    3.  Append new Phase to `plan.json` `phases`: `Phase X: Global Revision — [Change Topic] (<Date>)`, listing implementation tasks.
+    1.  Update `spec.md`.
+    2.  (UI projects only) Update `ui.md`; if screen structure affected, [[SKILL: archi-ui-wireframe|run skill (local update mode) to sync]][[NO-SKILL: (Skill not installed: read `[[__DOCS_DIR__]]/skills/archi-ui-wireframe/SKILL.md` and follow its protocol)]].
+    3.  Append new Phase to `plan.json`: `Phase X: Global Revision — [Change Topic] (<Date>)`.
 
-    **Output**: Change summary for each file (global + Task).
+    **Output**: Change summary for each file (global + Task). Enter step_5_verify.
 </step_4_execute>
 
-<step_5_summary>
-    **Role**: Auditor
-    **Checklist**:
-    1.  Cross-consistency between global assets (vision ↔ tech_stack ↔ roadmap ↔ map).
-    2.  Task docs aligned with updated global assets.
-    3.  No orphaned references (stale terms/paths in dictionary/map cleaned up).
+<step_5_verify>
+    **Role**: Independent Reviewer
 
-    **Terminal Gate** (Do not skip; must complete before output summary):
-    | Step | Command | Pass Condition |
-    |:---|:---|:---|
-    | 1 | `npx archi task --check` | No ERROR-level issues |
-    | 2 | `npx archi render` | `.md` views generated |
+    [[SUBAGENT: archi-silent-audit|mode: plan-docs, context: Review step_4 Phase 2 cascade-updated Task docs for alignment with modified global assets]][[NO-SKILL: (Skill not installed: read `[[__DOCS_DIR__]]/skills/archi-silent-audit/SKILL.md`, follow mode: plan-docs review dimension table)]]
 
-    **Action** (After Gate passes):
-    1.  Output change summary.
+    [[INCLUDE: shared/verify-result-handling.md]]
+</step_5_verify>
+
+<step_6_summary>
+    **Role**: Chief Auditor
 
-    **Output**:
-    ```
-    ## Global Revision Summary: [Change Topic]
+    **Pre-signoff Checklist** (confirm each item after Gate passes, before Output):
+    □ step_4 Safety Checkpoint — Git workspace confirmed (no uncommitted changes or user informed)
+    □ Phase 1 — all confirmed global assets modified, change summary output per file
+    □ (UI projects only, design system changed) User notified to re-run archi-ui-wireframe
+    □ Phase 2 — each affected Task: spec.md updated + plan.json appended Revision Phase
+    □ Step 5 Silent Audit — executed, all CRITICAL issues resolved
+    □ Consistency confirmed: vision ↔ tech_stack ↔ roadmap ↔ map, no orphan refs (stale terms/paths cleaned)
 
-    **Global Assets Modified**: [list of modified files]
-    **Tasks Updated**: [list of updated Tasks + impact summary each]
-    **Audit Result**: [pass/risk items]
+    **Terminal Gate** (do not skip): Standard check (task --check + render).
+
+    **Action** (after Gate passes):
+    1.  Output change summary.
 
-    ### Next Steps
-    | Priority | Action | Description |
-    |:---|:---|:---|
-    | ... | ... | ... |
-    ```
+    **Output**: Global Revision Summary — global asset changed files list, Task update list and impact summary, audit result, Next Steps table.
 </step_5_summary>
 
 </protocol_revise>