architext 0.0.3 → 0.0.5

package/dist/templates/en/skills/archi-feature-relations/SKILL.md

@@ -0,0 +1,118 @@
+---
+name: archi-feature-relations
+type: reviewer
+description: featureRelations linkage handler. In isolated context, handles map.json featureRelations register/check/cleanup, ensures aggregator Tasks maintain correct linkage with their sources.
+---
+
+# featureRelations Linkage Handler
+
+## System Flow Position
+
+```
+/archi.* step_N → Verify phase
+    ↓
+[This Skill] receives mode param → process featureRelations → return result
+    ↓
+Main Agent Signoff (confirm linkage prompts)
+```
+
+> **Skill responsibility boundary**:
+> - Responsible: featureRelations register/check/cleanup
+> - Not responsible: Modify business code, modify spec/plan docs
+
+---
+
+## Modes and Behavior
+
+### Mode `register` (caller: plan, inherit)
+
+Determine if current Task is an "aggregator Task"; if so, append to `map.json` featureRelations.
+
+**Aggregator Task criteria**:
+- Core responsibility is **listing, aggregating, or dynamically reflecting** other Tasks of a class
+- Code pattern: iterate/enumerate/dynamic load of similar modules (`for (const cmd of allCommands)`, `Object.values(registry)`, read dir then dynamic import)
+- Description pattern: "aggregate all page entries", "register all routes", "list all commands"
+
+**Action**:
+1. Analyze Task spec/goal description
+2. Determine if aggregator
+3. If yes → append featureRelations entry:
+   ```json
+   {
+     "aggregator": "<Task ID or file path>",
+     "sources": "<source scope description>",
+     "evidence": "<basis>",
+     "checkNote": "When such Tasks are added/removed, check if <aggregator> needs sync"
+   }
+   ```
+4. If no → output `NOT AGGREGATOR`, skip
+
+### Mode `check` (caller: scope, code)
+
+Compare current Task with featureRelations sources field semantically, output linkage prompts.
+
+**Action**:
+1. Read `map.json` featureRelations array
+2. If featureRelations empty → output `NO RELATIONS`, skip
+3. Compare this task/implemented feature with each sources semantically
+4. On match, output linkage prompt
+
+### Mode `cleanup` (caller: remove)
+
+Remove featureRelations entries referencing removed Task, assess impact.
+
+**Action**:
+1. Read `map.json` featureRelations array
+2. Find entries where aggregator or sources references removed Task
+3. Assess removal impact:
+   - Removed is aggregator → delete entire entry
+   - Removed is one of sources → update sources description, prompt to check if aggregator needs adjustment
+4. Output impact report
+
+---
+
+## Output Format
+
+### register mode
+
+```
+### Feature Relations: register
+
+RESULT: AGGREGATOR REGISTERED
+- aggregator: <ID>
+- sources: <description>
+- evidence: <basis>
+```
+
+Or: `RESULT: NOT AGGREGATOR — This Task is not aggregator, skip`
+
+### check mode
+
+```
+### Feature Relations: check
+
+HITS:
+- [aggregator] — [checkNote]
+
+NO HITS: [no matching relations]
+```
+
+Or: `NO RELATIONS — featureRelations is empty`
+
+### cleanup mode
+
+```
+### Feature Relations: cleanup
+
+REMOVED:
+- [deleted entries]
+
+UPDATED:
+- [entries with updated sources]
+
+IMPACT: [impact description]
+```
+
+---
+
+> **Intermediate output**: This Skill is a review subprogram; after producing result, control returns to caller.

package/dist/templates/en/skills/archi-interview-protocol/SKILL.md

@@ -1,5 +1,6 @@
 ---
 name: archi-interview-protocol
+type: specialist
 description: Architext supplementary interview protocol. Standardizes how to ask about information gaps: multiple-choice first, AI recommendation + [Recommended] marking, [Z] Custom fallback, complete AI+/AI- analysis, option descriptions state concrete behavior. Produces standard Q-table and INPUT prompt line, referenced by /archi.start, /archi.scope, and /archi.plan supplementary confirmation steps.
 ---
 
@@ -19,7 +20,7 @@ Information gap detected
 > - Responsible for: how to ask (format / rules / tone)
 > - Not responsible for: what to ask (determined by caller's gap list), how to handle user answers (determined by caller)
 
-## Callers & Trigger Conditions
+## Invocation Modes
 
 | Caller | Trigger Step | Trigger Condition | Max Questions |
 |:---|:---|:---|:---|

package/dist/templates/en/skills/archi-plan-options/SKILL.md

@@ -1,5 +1,6 @@
 ---
 name: archi-plan-options
+type: subprocess
 description: Architext architecture decision option library. Defines candidate approaches and AI+/AI- analysis for five core dimensions (Core Structure / Interaction Pattern / Data Flow / Error Handling / Access & Scope), covering Web/CLI/API/Lib/Mobile/MiniApp/Extension/Desktop/AI Agent project types. Includes convention inheritance rules, project tag routing logic, and recommend vs. expand criteria. Referenced by /archi.plan step_2 Part 2 (architecture recommendation phase).
 ---
 
@@ -14,7 +15,7 @@ description: Architext architecture decision option library. Defines candidate a
     ↓
 Direct recommendation row (most dimensions) or Q-table (when user decision needed)
     ↓
-Written into Feature Proposal architecture recommendation table
+Written into Task Proposal architecture recommendation table
 ```
 
 > **Skill responsibility boundary**:
@@ -36,7 +37,7 @@ Read `02_tech_stack.md` Section 9 (project conventions).
 
 ### Step 2 · Project Tag Routing
 
-Use project tags activated in step_1_load (`[?UI]` / `[?Data]` / `[?CLI]` / `[?Lib]` / `[?API]` / `[?Mobile]` / `[?MiniApp]` / `[?Extension]` / `[?Desktop]` / `[?AI]`) to select applicable dimensions; skip inapplicable ones. Routing rules are in each dimension's heading.
+Use project tags activated in step_1_load (`ui` / `data` / `cli` / `lib` / `api` / `mobile` / `miniapp` / `extension` / `desktop` / `ai`) to select applicable dimensions; skip inapplicable ones. Routing rules are in each dimension's heading.
 
 ### Step 3 · Recommend vs. Expand
 
@@ -361,4 +362,4 @@ Only add supplementary handling if this feature has **special exception scenario
 
 ---
 
-> **Intermediate artifact**: This Skill is a subroutine. After producing a recommendation row or Q-table, control returns to `/archi.plan` step_2, where the caller assembles the output into the Feature Proposal's architecture recommendation table.
+> **Intermediate artifact**: This Skill is a subroutine. After producing a recommendation row or Q-table, control returns to `/archi.plan` step_2, where the caller assembles the output into the Task Proposal's architecture recommendation table.

package/dist/templates/en/skills/archi-silent-audit/SKILL.md

@@ -0,0 +1,118 @@
+---
+name: archi-silent-audit
+type: reviewer
+description: Embedded lightweight review. In an isolated context, reviews main Agent output, filters review dimensions by mode, returns finding list. Must not generate report files. Shares dimension definitions with audit.md but differs in responsibility: audit.md is user-triggered standalone deep review; this Skill is protocol-triggered lightweight check.
+---
+
+# Embedded Lightweight Review
+
+## System Flow Position
+
+```
+/archi.* step_N → Verify phase
+    ↓
+[This Skill] receives mode param → isolated review → returns finding list
+    ↓
+Main Agent Signoff (must respond to findings)
+```
+
+> **Skill responsibility boundary**:
+> - Responsible: Review main Agent output in isolated context, output graded finding list
+> - Not responsible: Fix issues (return to main Agent), generate report files (that is audit.md's job), run Terminal Gate commands
+
+> **Relationship with `audit.md`**:
+> - `audit.md` = Standalone deep review protocol, generates report file, only user-triggered via `/archi.audit`
+> - This Skill = Embedded lightweight check, inline finding list, protocol-triggered
+> - Both share review dimension definitions (see `audit.md` step_2_task); this Skill filters by mode
+
+---
+
+## Modes and Review Dimensions
+
+### Mode `init` (caller: start, inherit)
+
+Review global file quality for new/inherited projects.
+
+| # | Dimension | Review Points |
+|:---|:---|:---|
+| 1 | **Vision-Roadmap alignment** | Roadmap task direction aligns with vision.md north star |
+| 2 | **Tech Stack consistency** | `02_tech_stack.md` matches actual deps/config |
+| 3 | **Global file completeness** | Required global files present (vision, roadmap, map, dictionary, tech_stack, custom_rules) |
+| 4 | **Zero info omission** | All Brief/code info routed to corresponding files |
+| 5 | [?UI] **Design Tokens** | `design_tokens.json` has base colors/fonts/spacing |
+
+### Mode `plan-docs` (caller: plan)
+
+Review planning doc (spec/ui/plan) quality.
+
+| # | Dimension | Review Points |
+|:---|:---|:---|
+| 1 | **Design Fidelity** | spec § 2 fully covers confirmed functional design |
+| 2 | **Dimension Match** | spec § 2 dimension format matches Task Type |
+| 3 | **Tech Consistency** | No tech not declared in `02_tech_stack.md` |
+| 4 | **WBS Coverage** | plan.json 100% covers each AC in spec |
+| 5 | **Notes Quality** | plan.json each task notes has deliverable+constraint+executable verification |
+| 6 | **Interface Exports** | INF task § 4 filled; interface declared when downstream deps exist |
+| 7 | **Constraints** | § 5 includes vision.md + tech_stack red lines |
+| 8 | [?Data] **Data Integrity** | Entities and fields match confirmed core entities |
+| 9 | [?Complex] **Design Trace** | design.md § 6 all ACs traceable |
+| 10 | [?Complex] **Parameter Specificity** | design.md § 3 params concrete (no vague terms) |
+| 11 | [?Complex] **Self-Check Pass** | design.md § 2 mechanism self-check list passes |
+
+### Mode `code-impl` (caller: code)
+
+Review code implementation quality.
+
+| # | Dimension | Review Points |
+|:---|:---|:---|
+| 1 | **Tech Consistency** | Matches `02_tech_stack.md` (libs/patterns/API style) |
+| 2 | **SOTA** | Reject outdated patterns; use tech_stack best practices |
+| 3 | **Security** | No sensitive info leak; input validated |
+| 4 | **Performance** | Avoid unnecessary large deps/full imports/useless computation/memory leaks |
+| 5 | [?UI] **Design Compliance** | Styles use Token/Preset only; no hardcoded magic values |
+| 6 | [?UI] **Accessibility** | Necessary a11y attributes present |
+| 7 | [?Data] **Data Integrity** | Matches `data_snapshot.json`; field names/types consistent |
+| 8 | [?i18n] **I18n** | No hardcoded strings; use Key/dictionary reference |
+| 9 | [?Complex] **Design Compliance** | State transitions/flows/protocols match design.md § 2 |
+| 10 | [?Complex] **Invariant Enforcement** | design.md § 4 invariants have assert/runtime checks in code |
+| 11 | [?Complex] **Parameter Alignment** | Code values match design.md § 3 param table |
+
+---
+
+## Execution Protocol
+
+1. **Load context**: Load required docs and code per caller-provided file paths
+2. **Filter dimensions by mode**: Execute only dimensions for current mode
+3. **Review item by item**: Each dimension outputs PASS or finding (level+location+description)
+4. **Output finding list**: Return sorted by level
+
+### Finding Levels
+
+| Level | Meaning | Main Agent must |
+|:---|:---|:---|
+| `CRITICAL` | Blocking issue | **Must fix** before signoff, cannot skip |
+| `WARNING` | Risk | **Must explain** handling in signoff report |
+| `INFO` | Suggestion | May decide whether to handle |
+
+### Output Format
+
+```
+### Silent Audit Results (mode: <mode>)
+
+**CRITICAL** (must fix):
+- [Dimension] location: description
+
+**WARNING** (must explain):
+- [Dimension] location: description
+
+**INFO** (suggestion):
+- [Dimension] location: description
+
+**Summary**: X CRITICAL / Y WARNING / Z INFO
+```
+
+When no findings: `### Silent Audit Results (mode: <mode>) — ALL PASS`
+
+---
+
+> **Intermediate output**: This Skill is a review subprogram; after producing finding list, control returns to caller; main Agent responds to findings in Signoff.