invar-tools 1.12.0__py3-none-any.whl → 1.14.0__py3-none-any.whl

invar/templates/protocol/python/tools.md

@@ -1,9 +1,9 @@
 ## Commands (Python)
 
 ```bash
-invar guard              # Full: static + doctests + CrossHair + Hypothesis
+invar guard              # Check git-modified files (fast, default)
+invar guard --all        # Check entire project (CI, release)
 invar guard --static     # Static only (quick debug, ~0.5s)
-invar guard --changed    # Modified files only
 invar guard --coverage   # Collect branch coverage
 invar guard -c           # Contract coverage only (DX-63)
 invar sig <file>         # Show contracts + signatures
@@ -11,6 +11,9 @@ invar map --top 10       # Most-referenced symbols
 invar rules              # List all rules with detection/hints (JSON)
 ```
 
+**Default behavior**: Checks git-modified files for fast feedback during development.
+Use `--all` for comprehensive checks before release.
+
 ## Configuration (Python)
 
 ```toml

invar/templates/protocol/typescript/tools.md

@@ -2,15 +2,18 @@
 
 ```bash
 # Verification (Python CLI - works for TypeScript)
-invar guard                  # Full: tsc + eslint + vitest + ts-analyzer
+invar guard                  # Check git-modified files (fast, default)
+invar guard --all            # Check entire project (CI, release)
 invar guard --json           # Agent-friendly v2.0 JSON output
-invar guard --changed        # Modified files only
 
 # Analysis
 invar sig <file>             # Show function signatures
 invar map --top 10           # Most-referenced symbols
 ```
 
+**Default behavior**: Checks git-modified files for fast feedback during development.
+Use `--all` for comprehensive checks before release.
+
 ## Guard Output (v2.0 JSON)
 
 ```json

invar/templates/skills/invar-reflect/CONFIG.md

@@ -0,0 +1,355 @@
+# /invar-reflect Configuration
+
+**Status**: Phase B - Hook Integration (Proposed)
+
+This document describes the configuration for automatic feedback collection using the `/invar-reflect` skill.
+
+---
+
+## Overview
+
+The `/invar-reflect` skill can be triggered:
+1. **Manually**: User calls `/invar-reflect`
+2. **Automatically** (Phase B): Via `PostTaskCompletion` hook when conditions are met
+
+---
+
+## Proposed Hook Schema
+
+### PostTaskCompletion Hook (Waiting for Claude Code Support)
+
+**Proposed configuration in `.claude/settings.json`**:
+
+```json
+{
+  "hooks": {
+    "PostTaskCompletion": [
+      {
+        "hooks": [
+          {
+            "type": "skill",
+            "skill": "invar-reflect",
+            "mode": "silent",
+            "conditions": {
+              "min_messages": 30,
+              "min_duration_hours": 2
+            }
+          }
+        ]
+      }
+    ]
+  },
+  "feedback": {
+    "enabled": true,
+    "auto_trigger": true,
+    "retention_days": 90
+  }
+}
+```
+
+**Hook Parameters**:
+
+| Parameter | Type | Default | Description |
+|-----------|------|---------|-------------|
+| `type` | string | - | Must be `"skill"` |
+| `skill` | string | - | Skill name: `"invar-reflect"` |
+| `mode` | string | `"silent"` | Silent mode (no user interruption) |
+| `conditions.min_messages` | number | 30 | Minimum messages in session |
+| `conditions.min_duration_hours` | number | 2 | Minimum session duration |
+
+**Feedback Config**:
+
+| Parameter | Type | Default | Description |
+|-----------|------|---------|-------------|
+| `enabled` | boolean | true | Enable feedback collection |
+| `auto_trigger` | boolean | true | Auto-run via hook |
+| `retention_days` | number | 90 | Auto-cleanup old files |
+
+---
+
+## Triggering Conditions
+
+The hook triggers when **ALL** conditions are met:
+
+1. ✅ **Task completed** - User finished major work (natural stopping point)
+2. ✅ **Message count >= 30** - Sufficient context for meaningful feedback
+3. ✅ **Duration >= 2 hours** - Non-trivial session (avoids quick fixes)
+
+**No hard frequency cap**: Same-day sessions merge into single file (see SKILL.md for merge logic).
+
+---
+
+## Silent Mode
+
+When `mode: "silent"`:
+- Feedback generation runs in background
+- No interruption to current conversation
+- User sees notification only: `✓ Feedback saved to .invar/feedback/feedback-{date}.md`
+
+---
+
+## User Control
+
+### Enable/Disable
+
+**During init** (Phase C):
+```bash
+$ invar init
+...
+Enable automatic feedback collection? [Y/n]: n
+```
+
+**After init**:
+```json
+// .claude/settings.json
+{
+  "feedback": {
+    "enabled": false  // Disable feedback collection
+  }
+}
+```
+
+Or disable auto-trigger only (manual `/invar-reflect` still works):
+```json
+{
+  "feedback": {
+    "enabled": true,
+    "auto_trigger": false  // Manual only
+  }
+}
+```
+
+### File Management
+
+Feedback files location:
+```
+.invar/feedback/
+├── feedback-2026-01-03.md  # All sessions from Jan 3
+├── feedback-2026-01-04.md  # All sessions from Jan 4
+└── feedback-2026-01-10.md  # Jan 10 (gaps are OK)
+```
+
+**Auto-cleanup**: Files older than `retention_days` are automatically deleted.
+
+**Manual cleanup**:
+```bash
+# Delete all feedback
+rm -rf .invar/feedback/
+
+# Delete specific file
+rm .invar/feedback/feedback-2026-01-03.md
+```
+
+---
+
+## Workaround: Using Stop Hook (Until PostTaskCompletion is Available)
+
+Since Claude Code doesn't yet support `PostTaskCompletion` hook, you can use the `Stop` hook as a temporary workaround.
+
+### Stop Hook Implementation
+
+Create `.claude/hooks/invar.FeedbackTrigger.sh`:
+
+```bash
+#!/bin/bash
+# Invar Feedback Trigger (Stop Hook Workaround)
+# DX-79 Phase B: Auto-trigger /invar-reflect on session end
+
+# Read feedback config
+FEEDBACK_ENABLED=$(jq -r '.feedback.enabled // true' .claude/settings.local.json 2>/dev/null)
+AUTO_TRIGGER=$(jq -r '.feedback.auto_trigger // true' .claude/settings.local.json 2>/dev/null)
+
+if [[ "$FEEDBACK_ENABLED" != "true" ]] || [[ "$AUTO_TRIGGER" != "true" ]]; then
+  exit 0  # Feedback disabled
+fi
+
+# Check session conditions (requires session state tracking)
+STATE_DIR="${CLAUDE_STATE_DIR:-/tmp/invar_hooks_$(id -u)}"
+SESSION_START_FILE="$STATE_DIR/session_start"
+MESSAGE_COUNT_FILE="$STATE_DIR/message_count"
+
+# Calculate duration
+if [[ -f "$SESSION_START_FILE" ]]; then
+  SESSION_START=$(cat "$SESSION_START_FILE")
+  SESSION_END=$(date +%s)
+  DURATION_HOURS=$(( (SESSION_END - SESSION_START) / 3600 ))
+else
+  DURATION_HOURS=0
+fi
+
+# Read message count
+MESSAGE_COUNT=$(cat "$MESSAGE_COUNT_FILE" 2>/dev/null || echo 0)
+
+# Check conditions
+if [[ $MESSAGE_COUNT -ge 30 ]] && [[ $DURATION_HOURS -ge 2 ]]; then
+  echo ""
+  echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+  echo "📊 Invar: Generating usage feedback..."
+  echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+  echo ""
+  echo "Session complete: $MESSAGE_COUNT messages, ${DURATION_HOURS}h duration"
+  echo "Feedback will be saved to .invar/feedback/feedback-$(date +%Y-%m-%d).md"
+  echo ""
+  echo "To review or share: cat .invar/feedback/feedback-$(date +%Y-%m-%d).md"
+  echo "To disable: Set feedback.auto_trigger=false in .claude/settings.local.json"
+fi
+
+# Cleanup session state
+rm -rf "$STATE_DIR" 2>/dev/null
+```
+
+**Note**: This workaround provides notification but doesn't actually generate feedback automatically. Full auto-generation requires:
+1. Claude Code to support skill invocation from hooks, OR
+2. Hook to write state file that triggers `/invar-reflect` on next user interaction
+
+---
+
+## Configuration Template (for `invar init`)
+
+When Phase C is complete, `invar init` will generate this configuration:
+
+```json
+{
+  "permissions": {
+    "allow": [
+      // ... existing permissions ...
+    ]
+  },
+  "hooks": {
+    "PostTaskCompletion": [
+      {
+        "hooks": [
+          {
+            "type": "skill",
+            "skill": "invar-reflect",
+            "mode": "silent",
+            "conditions": {
+              "min_messages": 30,
+              "min_duration_hours": 2
+            }
+          }
+        ]
+      }
+    ],
+    // ... existing hooks ...
+  },
+  "feedback": {
+    "enabled": true,
+    "auto_trigger": true,
+    "retention_days": 90
+  }
+}
+```
+
+**Init prompt** (Phase C):
+
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+📊 Usage Feedback (Optional)
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+Invar can automatically reflect on tool usage to help improve
+the framework. Feedback is:
+  - Stored locally in .invar/feedback/
+  - Never sent automatically
+  - You decide what (if anything) to share
+
+Enable automatic feedback collection? [Y/n]:
+```
+
+---
+
+## Privacy & Security
+
+**What's stored**:
+- Tool usage statistics (counts, success/failure)
+- Error types (NO error messages)
+- Session metadata (duration, message count)
+- User's subjective experience (generated by AI)
+
+**What's NOT stored**:
+- Source code
+- File paths (anonymized)
+- Error messages (might contain code)
+- Project-specific details
+
+**All data stays local** in `.invar/feedback/`. User reviews before sharing.
+
+---
+
+## Testing (Phase B Acceptance Criteria)
+
+### Manual Testing
+
+1. **Test manual invocation**:
+   ```bash
+   # In Claude Code session
+   /invar-reflect
+   ```
+   - ✅ Generates feedback file
+   - ✅ Saves to `.invar/feedback/feedback-{today}.md`
+   - ✅ No errors
+
+2. **Test config disable**:
+   ```json
+   // .claude/settings.local.json
+   { "feedback": { "enabled": false } }
+   ```
+   - ✅ `/invar-reflect` still works (manual override)
+   - ✅ Hook doesn't trigger (when implemented)
+
+3. **Test same-day merge**:
+   ```bash
+   # Session 1: Morning
+   /invar-reflect
+   # Creates feedback-2026-01-03.md
+
+   # Session 2: Afternoon (same day)
+   /invar-reflect
+   # Updates feedback-2026-01-03.md (not create new file)
+   ```
+   - ✅ Single file per day
+   - ✅ Intelligent merging (semantic understanding)
+   - ✅ Daily summary regenerated
+
+### Automated Testing (when hook is implemented)
+
+1. **Test condition thresholds**:
+   - Session < 30 messages: Hook doesn't trigger
+   - Session < 2 hours: Hook doesn't trigger
+   - Session >= 30 messages AND >= 2 hours: Hook triggers
+
+2. **Test silent mode**:
+   - Hook triggers in background
+   - No interruption to conversation
+   - Notification shown only
+
+3. **Test retention**:
+   - Files older than 90 days auto-deleted
+   - Configurable via `retention_days`
+
+---
+
+## Phase B Status
+
+**Completed**:
+- ✅ Hook schema designed
+- ✅ Configuration structure defined
+- ✅ User control mechanism specified
+- ✅ Stop hook workaround documented
+
+**Waiting for Claude Code Support**:
+- ⏸️ PostTaskCompletion hook type
+- ⏸️ Skill invocation from hooks
+- ⏸️ Session state tracking (message count, duration)
+
+**Next Steps**:
+- Implement Phase C: Init Integration (can be done independently)
+- Submit feature request to Claude Code for PostTaskCompletion hook
+- Test manual `/invar-reflect` invocation thoroughly
+
+---
+
+**Version**: 1.0 (Phase B - Proposed)
+**Updated**: 2026-01-03
+**Related**: DX-79 Invar Usage Feedback Collection