specweave 1.0.98 → 1.0.99

package/CLAUDE.md

@@ -1,10 +1,10 @@
-<!-- SW:META template="claude" version="1.0.97" sections="header,start,autodetect,metarule,rules,workflow,reflect,context,lsp,structure,taskformat,secrets,syncing,mapping,testing,api,limits,troubleshooting,principles,linking,mcp,autoexecute,auto,docs" -->
+<!-- SW:META template="claude" version="1.0.98" sections="header,start,autodetect,metarule,rules,workflow,reflect,context,lsp,structure,taskformat,secrets,syncing,mapping,testing,api,limits,troubleshooting,principles,linking,mcp,autoexecute,auto,docs" -->
 
-<!-- SW:SECTION:header version="1.0.97" -->
+<!-- SW:SECTION:header version="1.0.98" -->
 **Framework**: SpecWeave | **Truth**: `spec.md` + `tasks.md`
 <!-- SW:END:header -->
 
-<!-- SW:SECTION:start version="1.0.97" -->
+<!-- SW:SECTION:start version="1.0.98" -->
 ## Getting Started
 
 **Initial increment**: `0001-project-setup` (auto-created by `specweave init`)
@@ -14,7 +14,7 @@
 2. **Customize**: Edit spec.md and use for setup tasks
 <!-- SW:END:start -->
 
-<!-- SW:SECTION:autodetect version="1.0.97" -->
+<!-- SW:SECTION:autodetect version="1.0.98" -->
 ## Auto-Detection
 
 SpecWeave auto-detects product descriptions and routes to `/sw:increment`:
@@ -24,7 +24,7 @@ SpecWeave auto-detects product descriptions and routes to `/sw:increment`:
 **Opt-out phrases**: "Just brainstorm first" | "Don't plan yet" | "Quick discussion" | "Let's explore ideas"
 <!-- SW:END:autodetect -->
 
-<!-- SW:SECTION:metarule version="1.0.97" -->
+<!-- SW:SECTION:metarule version="1.0.98" -->
 ## Meta-Rule: Think-Before-Act
 
 **Satisfy dependencies BEFORE dependent operations.**
@@ -35,7 +35,7 @@ SpecWeave auto-detects product descriptions and routes to `/sw:increment`:
 ```
 <!-- SW:END:metarule -->
 
-<!-- SW:SECTION:rules version="1.0.97" -->
+<!-- SW:SECTION:rules version="1.0.98" -->
 ## Rules
 
 1. **Files** → `.specweave/increments/####-name/` (spec.md, plan.md, tasks.md at root; reports/, scripts/, logs/ subfolders)
@@ -47,7 +47,7 @@ SpecWeave auto-detects product descriptions and routes to `/sw:increment`:
 7. **⛔ Initialization guard**: `.specweave/` folders MUST ONLY exist where `specweave init` was run. NEVER create `.specweave/` in parent, nested, or unrelated directories. Check `config.json` exists before creating ANY `.specweave/` subfolders.
 <!-- SW:END:rules -->
 
-<!-- SW:SECTION:workflow version="1.0.97" -->
+<!-- SW:SECTION:workflow version="1.0.98" -->
 ## Workflow
 
 `/sw:increment "X"` → `/sw:do` → `/sw:progress` → `/sw:done 0001`
@@ -67,7 +67,7 @@ SpecWeave auto-detects product descriptions and routes to `/sw:increment`:
 **Natural language**: "Let's build X" → `/sw:increment` | "What's status?" → `/sw:progress` | "We're done" → `/sw:done` | "Ship while sleeping" → `/sw:auto`
 <!-- SW:END:workflow -->
 
-<!-- SW:SECTION:reflect version="1.0.97" -->
+<!-- SW:SECTION:reflect version="1.0.98" -->
 ## Self-Improving Skills (Reflect)
 
 **Learn once, never repeat.** Claude learns from corrections and patterns across sessions.
@@ -111,7 +111,7 @@ ls ~/.specweave/memory/*.md 2>/dev/null
 **Enable auto-learning**: `/sw:reflect-on` → Stop hook analyzes sessions automatically
 <!-- SW:END:reflect -->
 
-<!-- SW:SECTION:context version="1.0.97" -->
+<!-- SW:SECTION:context version="1.0.98" -->
 ## Living Docs Context
 
 **Before implementing features**: Check existing docs for patterns and decisions.
@@ -131,7 +131,7 @@ grep -ril "keyword" .specweave/docs/internal/
 **Use `/sw:context <topic>`** to load relevant living docs into conversation.
 <!-- SW:END:context -->
 
-<!-- SW:SECTION:lsp version="1.0.97" -->
+<!-- SW:SECTION:lsp version="1.0.98" -->
 ## LSP-Enhanced Exploration
 
 **USE LSP ACTIVELY** for semantic code understanding (100x faster than grep).
@@ -148,7 +148,7 @@ go install golang.org/x/tools/gopls@latest  # Go
 **Best Practices**: ALWAYS use `findReferences` before refactoring | Use `goToDefinition` instead of grep | Combine with Explore agent
 <!-- SW:END:lsp -->
 
-<!-- SW:SECTION:structure version="1.0.97" -->
+<!-- SW:SECTION:structure version="1.0.98" -->
 ## Structure
 
 ```
@@ -218,7 +218,7 @@ my-project/
 ```
 <!-- SW:END:structure -->
 
-<!-- SW:SECTION:taskformat version="1.0.97" -->
+<!-- SW:SECTION:taskformat version="1.0.98" -->
 ## Task Format
 
 ```markdown
@@ -228,7 +228,7 @@ my-project/
 ```
 <!-- SW:END:taskformat -->
 
-<!-- SW:SECTION:secrets version="1.0.97" -->
+<!-- SW:SECTION:secrets version="1.0.98" -->
 ## Secrets Check
 
 **BEFORE CLI tools**: Check existing config first!
@@ -239,7 +239,7 @@ gh auth status
 ```
 <!-- SW:END:secrets -->
 
-<!-- SW:SECTION:syncing version="1.0.97" -->
+<!-- SW:SECTION:syncing version="1.0.98" -->
 ## External Sync (GitHub/JIRA/ADO)
 
 **After increment creation**: Run `/sw-github:sync {id}` to create issues!
@@ -267,7 +267,7 @@ Living docs sync ≠ External sync. They are separate:
 **Verify tokens**: `grep GITHUB_TOKEN .env` | `gh auth status`
 <!-- SW:END:syncing -->
 
-<!-- SW:SECTION:mapping version="1.0.97" -->
+<!-- SW:SECTION:mapping version="1.0.98" -->
 ## GitHub Mapping
 
 | SpecWeave | GitHub |
@@ -277,7 +277,7 @@ Living docs sync ≠ External sync. They are separate:
 | Task T-XXX | Checkbox |
 <!-- SW:END:mapping -->
 
-<!-- SW:SECTION:testing version="1.0.97" -->
+<!-- SW:SECTION:testing version="1.0.98" -->
 ## Testing
 
 BDD in tasks.md | Unit >80% | `.test.ts` (Vitest)
@@ -289,7 +289,7 @@ vi.mock('fs', () => ({ readFile: vi.fn() }));
 ```
 <!-- SW:END:testing -->
 
-<!-- SW:SECTION:api version="1.0.97" -->
+<!-- SW:SECTION:api version="1.0.98" -->
 ## API Development (OpenAPI-First)
 
 **For API projects only.** OpenAPI = source of truth → Postman derived from it.
@@ -308,13 +308,13 @@ vi.mock('fs', () => ({ readFile: vi.fn() }));
 **Import**: Postman → Import collection + env → Fill secrets → Select env
 <!-- SW:END:api -->
 
-<!-- SW:SECTION:limits version="1.0.97" -->
+<!-- SW:SECTION:limits version="1.0.98" -->
 ## Limits
 
 **Max 1500 lines/file** — extract before adding
 <!-- SW:END:limits -->
 
-<!-- SW:SECTION:troubleshooting version="1.0.97" -->
+<!-- SW:SECTION:troubleshooting version="1.0.98" -->
 ## Troubleshooting
 
 | Issue | Fix |
@@ -333,7 +333,7 @@ vi.mock('fs', () => ({ readFile: vi.fn() }));
 | Path patterns not working | `//path` = absolute, `/path` = relative to settings file, `additionalDirectories` for explicit working dirs |
 <!-- SW:END:troubleshooting -->
 
-<!-- SW:SECTION:principles version="1.0.97" -->
+<!-- SW:SECTION:principles version="1.0.98" -->
 ## Principles
 
 1. **Spec-first**: `/sw:increment` before coding
@@ -343,7 +343,7 @@ vi.mock('fs', () => ({ readFile: vi.fn() }));
 5. **Clean**: All files in increment folders
 <!-- SW:END:principles -->
 
-<!-- SW:SECTION:linking version="1.0.97" -->
+<!-- SW:SECTION:linking version="1.0.98" -->
 ## Bidirectional Linking
 
 Tasks ↔ User Stories auto-linked via AC-IDs: `AC-US1-01` → `US-001`
@@ -351,7 +351,7 @@ Tasks ↔ User Stories auto-linked via AC-IDs: `AC-US1-01` → `US-001`
 Task format: `**AC**: AC-US1-01, AC-US1-02` (CRITICAL for linking)
 <!-- SW:END:linking -->
 
-<!-- SW:SECTION:mcp version="1.0.97" -->
+<!-- SW:SECTION:mcp version="1.0.98" -->
 ## External Service Connection
 
 **Priority**: MCP Server → REST API → CLI → Direct Connection
@@ -375,7 +375,7 @@ wrangler whoami 2>/dev/null
 ```
 <!-- SW:END:mcp -->
 
-<!-- SW:SECTION:autoexecute version="1.0.97" -->
+<!-- SW:SECTION:autoexecute version="1.0.98" -->
 ## Auto-Execute Rule
 
 **NEVER** output "Manual Step Required" when credentials exist. **EXECUTE DIRECTLY.**
@@ -389,7 +389,7 @@ wrangler whoami 2>/dev/null && gh auth status 2>/dev/null
 ```
 <!-- SW:END:autoexecute -->
 
-<!-- SW:SECTION:auto version="1.0.97" -->
+<!-- SW:SECTION:auto version="1.0.98" -->
 ## Auto Mode (Autonomous Execution)
 
 **Continuous execution until all tasks complete.**
@@ -466,7 +466,7 @@ wrangler whoami 2>/dev/null && gh auth status 2>/dev/null
 **Circuit Breaker**: External API fails 3x? Queue & continue
 <!-- SW:END:auto -->
 
-<!-- SW:SECTION:docs version="1.0.97" -->
+<!-- SW:SECTION:docs version="1.0.98" -->
 ## Docs
 
 [spec-weave.com](https://spec-weave.com) | `.specweave/docs/internal/`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "specweave",
-  "version": "1.0.98",
+  "version": "1.0.99",
   "description": "Spec-driven development framework for Claude Code. AI-native workflow with living documentation, intelligent agents, and multilingual support (9 languages). Enterprise-grade traceability with permanent specs and temporary increments.",
   "type": "module",
   "main": "dist/index.js",

package/plugins/specweave/scripts/reflect.sh

@@ -123,30 +123,84 @@ is_actionable() {
     # Must contain some actionable verb
     echo "$text" | grep -qiE "(use|don.t|never|always|should|must|avoid|prefer)" || return 1
 
+    # QUALITY GATE: Skip documentation examples (markdown formatting, line numbers, JSON)
+    # These patterns indicate the text came from docs/examples, not actual user corrections
+    if echo "$text" | grep -qE '^\s*[0-9]+→|^\s*[0-9]+\s*\||```|\*\*|\\n|\\"|":[[:space:]]|"sessionId"|"userType"|"cwd"'; then
+        return 1
+    fi
+
+    # QUALITY GATE: Skip content that looks like code examples with backticks
+    if echo "$text" | grep -qE '`[^`]+`.*`[^`]+`'; then
+        return 1
+    fi
+
+    # QUALITY GATE: Skip content with escaped quotes (JSON/code artifacts)
+    if echo "$text" | grep -qE '\\"|\\".*\\"'; then
+        return 1
+    fi
+
     return 0
 }
 
 # Extract the actual rule from a correction context
+# PRIORITY: Extract the "do Y instead" part, not the "don't X" part
 extract_rule() {
     local context="$1"
-
-    # Try to find the "do Y instead" part
     local rule=""
 
-    # Pattern: "No, don't X. Use Y instead" -> "Use Y"
-    rule=$(echo "$context" | grep -oiE "(use|prefer|always)[^.!?]*" | head -1)
-    [ -n "$rule" ] && echo "$rule" && return
+    # PRIORITY 1: Look for "Use X instead" pattern
+    # Captures "Use logger.info() instead" - allows alphanumeric, dots, parens before "instead"
+    rule=$(echo "$context" | grep -oiE "[Uu]se [a-zA-Z0-9_.()/]+(\s+[a-zA-Z0-9_.()/]+)*\s+instead" | head -1)
+    if [ -n "$rule" ] && [ ${#rule} -gt 15 ]; then
+        echo "$rule" | sed 's/^[[:space:]]*//' | cut -c1-100
+        return
+    fi
+
+    # PRIORITY 2: Look for "should use X" (correction pattern)
+    rule=$(echo "$context" | grep -oiE "should (use|be) [^.!?]+" | head -1)
+    if [ -n "$rule" ] && [ ${#rule} -gt 15 ]; then
+        echo "$rule" | sed 's/^[[:space:]]*//' | cut -c1-100
+        return
+    fi
+
+    # PRIORITY 3: Look for sentence after "Wrong!" or "No,"
+    # Extract the sentence that comes after the negation marker
+    rule=$(echo "$context" | grep -oiE "(Wrong!?|No,) [^.!?]+" | sed 's/^[Ww]rong!*[[:space:]]*//;s/^[Nn]o,[[:space:]]*//' | head -1)
+    if [ -n "$rule" ] && [ ${#rule} -gt 15 ]; then
+        echo "$rule" | sed 's/^[[:space:]]*//' | cut -c1-100
+        return
+    fi
+
+    # PRIORITY 4: "Always X" (explicit rule)
+    rule=$(echo "$context" | grep -oiE "[Aa]lways [^.!?]+" | head -1)
+    if [ -n "$rule" ] && [ ${#rule} -gt 15 ]; then
+        echo "$rule" | sed 's/^[[:space:]]*//' | cut -c1-100
+        return
+    fi
+
+    # PRIORITY 5: "Never X" (prohibition rule)
+    rule=$(echo "$context" | grep -oiE "[Nn]ever [^.!?]+" | head -1)
+    if [ -n "$rule" ] && [ ${#rule} -gt 15 ]; then
+        echo "$rule" | sed 's/^[[:space:]]*//' | cut -c1-100
+        return
+    fi
 
-    # Pattern: "Should be X" -> "Should be X"
-    rule=$(echo "$context" | grep -oiE "should (be|use)[^.!?]*" | head -1)
-    [ -n "$rule" ] && echo "$rule" && return
+    # PRIORITY 6: "In this project/codebase, use X"
+    rule=$(echo "$context" | grep -oiE "[Ii]n this (project|codebase)[^.!?]+" | head -1)
+    if [ -n "$rule" ] && [ ${#rule} -gt 15 ]; then
+        echo "$rule" | sed 's/^[[:space:]]*//' | cut -c1-100
+        return
+    fi
 
-    # Pattern: "Never X, always Y" -> "Always Y"
-    rule=$(echo "$context" | grep -oiE "always[^.!?]*" | head -1)
-    [ -n "$rule" ] && echo "$rule" && return
+    # PRIORITY 7: "prefer X" or "use X not Y"
+    rule=$(echo "$context" | grep -oiE "(prefer|use) [^.!?]+ (not|instead|over) [^.!?]+" | head -1)
+    if [ -n "$rule" ] && [ ${#rule} -gt 15 ]; then
+        echo "$rule" | sed 's/^[[:space:]]*//' | cut -c1-100
+        return
+    fi
 
-    # Fallback: return cleaned context
-    echo "$context" | cut -c1-100
+    # Fallback: return cleaned context (last resort, but apply quality filter in add_rule)
+    echo "$context" | sed 's/^[[:space:]]*//;s/^[Uu]ser:[[:space:]]*//' | cut -c1-100
 }
 
 # Detect high-value signals in transcript
@@ -205,25 +259,51 @@ detect_signals() {
 # DEDUPLICATION
 # ============================================================================
 
-# Check if a similar rule already exists (by key words)
+# Check if a similar rule already exists
+# STRICT DEDUPLICATION: Exact substring match OR high keyword overlap
 rule_exists() {
     local file="$1"
     local new_rule="$2"
 
     [ ! -f "$file" ] && return 1
 
-    # Extract key words from new rule (nouns, verbs)
-    local keywords=$(echo "$new_rule" | tr '[:upper:]' '[:lower:]' | grep -oE '\b[a-z]{4,}\b' | sort -u | head -5 | tr '\n' ' ')
+    # Normalize rule for comparison (lowercase, trim, collapse spaces)
+    local normalized=$(echo "$new_rule" | tr '[:upper:]' '[:lower:]' | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/[[:space:]]\+/ /g')
 
-    # Check if 3+ keywords match any existing rule
-    local match_count=0
-    for kw in $keywords; do
-        if grep -qi "\b$kw\b" "$file" 2>/dev/null; then
-            match_count=$((match_count + 1))
-        fi
-    done
+    # STRICT CHECK 1: Exact substring match (catches same rule with different prefix)
+    if grep -qi "$normalized" "$file" 2>/dev/null; then
+        return 0
+    fi
+
+    # STRICT CHECK 2: Rule already contains the key phrase
+    # Extract core action phrase (verb + object)
+    local core_phrase=$(echo "$normalized" | grep -oE "(use|prefer|always|never) [a-z]+( [a-z]+)?" | head -1)
+    if [ -n "$core_phrase" ] && grep -qi "$core_phrase" "$file" 2>/dev/null; then
+        return 0
+    fi
+
+    # STRICT CHECK 3: High keyword overlap (at least 50% of words must match a single existing rule)
+    local keywords=$(echo "$normalized" | grep -oE '\b[a-z]{4,}\b' | sort -u | tr '\n' ' ')
+    local keyword_count=$(echo "$keywords" | wc -w | tr -d ' ')
+
+    # For each existing rule, check overlap
+    while IFS= read -r existing_rule; do
+        [ -z "$existing_rule" ] && continue
+        local existing_lower=$(echo "$existing_rule" | tr '[:upper:]' '[:lower:]')
+        local match_count=0
+
+        for kw in $keywords; do
+            if echo "$existing_lower" | grep -q "\b$kw\b" 2>/dev/null; then
+                match_count=$((match_count + 1))
+            fi
+        done
+
+        # If more than 50% of keywords match ONE existing rule, it's a duplicate
+        local threshold=$((keyword_count / 2))
+        [ "$threshold" -lt 2 ] && threshold=2
+        [ "$match_count" -ge "$threshold" ] && return 0
+    done < <(grep "^- " "$file" 2>/dev/null)
 
-    [ "$match_count" -ge 3 ] && return 0
     return 1
 }
 
@@ -236,13 +316,16 @@ categorize() {
     local lower=$(echo "$text" | tr '[:upper:]' '[:lower:]')
 
     case "$lower" in
-        *component*|*button*|*ui*|*style*|*css*|*react*|*vue*|*html*) echo "frontend" ;;
-        *api*|*endpoint*|*route*|*rest*|*graphql*|*server*) echo "backend" ;;
-        *test*|*spec*|*mock*|*assert*|*expect*) echo "testing" ;;
-        *deploy*|*docker*|*k8s*|*ci*|*terraform*|*aws*) echo "devops" ;;
-        *auth*|*security*|*token*|*password*|*secret*) echo "security" ;;
-        *query*|*database*|*sql*|*schema*|*table*) echo "database" ;;
-        *file*|*path*|*import*|*export*|*module*) echo "structure" ;;
+        *component*|*button*|*ui*|*style*|*css*|*react*|*vue*|*html*|*tailwind*) echo "frontend" ;;
+        *api*|*endpoint*|*route*|*rest*|*graphql*|*server*|*backend*) echo "backend" ;;
+        *test*|*spec*|*mock*|*assert*|*expect*|*vitest*|*jest*|*playwright*) echo "testing" ;;
+        *deploy*|*docker*|*k8s*|*ci*|*terraform*|*aws*|*gcp*|*azure*) echo "devops" ;;
+        *auth*|*security*|*token*|*password*|*secret*|*encryption*) echo "security" ;;
+        *query*|*database*|*sql*|*schema*|*table*|*prisma*|*drizzle*) echo "database" ;;
+        *file*|*path*|*import*|*export*|*module*|*require*) echo "structure" ;;
+        *logger*|*log*|*console*|*debug*|*error*|*warn*) echo "logging" ;;
+        *type*|*interface*|*typescript*|*generic*|*enum*) echo "types" ;;
+        *git*|*commit*|*branch*|*merge*|*rebase*) echo "git" ;;
         *) echo "general" ;;
     esac
 }
@@ -273,14 +356,42 @@ add_rule() {
 
     local file=$(ensure_memory_file "$category")
 
+    # Clean and format FIRST
+    # Remove "User:" prefix, extra whitespace, and limit length
+    local clean=$(echo "$rule" | tr -d '\n\r' | sed 's/  */ /g' | sed 's/^[[:space:]]*//;s/[[:space:]]*$//' | sed 's/^[Uu]ser:[[:space:]]*//' | cut -c1-100)
+
+    # QUALITY GATE 1: Minimum length (too short = useless)
+    if [ ${#clean} -lt 15 ]; then
+        log "info" "Skipped too short: $clean"
+        return 1
+    fi
+
+    # QUALITY GATE 2: Must contain actionable verb
+    if ! echo "$clean" | grep -qiE "(use|prefer|always|never|should|avoid|don't|must|do not)"; then
+        log "info" "Skipped non-actionable: $clean"
+        return 1
+    fi
+
+    # QUALITY GATE 3: Must have specific subject (not just a verb phrase)
+    local word_count=$(echo "$clean" | wc -w | tr -d ' ')
+    if [ "$word_count" -lt 3 ]; then
+        log "info" "Skipped too vague: $clean"
+        return 1
+    fi
+
+    # QUALITY GATE 4: Skip documentation/JSON artifacts
+    # These patterns indicate corrupted data from parsing docs or code
+    if echo "$clean" | grep -qE '\\n|\\"|":|→|```|\*\*|sessionId|userType|"cwd"'; then
+        log "info" "Skipped doc/JSON artifact: $clean"
+        return 1
+    fi
+
     # Skip if similar rule exists
-    if rule_exists "$file" "$rule"; then
-        log "info" "Skipped duplicate: $rule"
+    if rule_exists "$file" "$clean"; then
+        log "info" "Skipped duplicate: $clean"
         return 1
     fi
 
-    # Clean and format
-    local clean=$(echo "$rule" | tr -d '\n\r' | sed 's/  */ /g' | cut -c1-100)
     local marker="→"
     [ "$type" = "CORRECTION" ] && marker="✗→✓"
 

package/plugins/specweave/skills/update-instructions/SKILL.md

@@ -17,8 +17,8 @@ Smart merge for CLAUDE.md and AGENTS.md instruction files.
 
 ## When to Use
 
-- After running `bash scripts/refresh-marketplace.sh`
-- After upgrading SpecWeave version (`npm update specweave`)
+- After running `specweave refresh-marketplace`
+- After upgrading SpecWeave version (`npm update -g specweave`)
 - When CLAUDE.md or AGENTS.md seem outdated
 - To sync instruction files with latest framework features