specky-sdd 3.4.0-rc.8 → 3.4.0

package/.apm/agents/design-architect.agent.md

@@ -2,8 +2,6 @@
 name: design-architect
 description: Phase 4 agent that writes DESIGN.md with system architecture, API contracts, data models, and Mermaid diagrams. All design decisions trace to specification requirements.
 
-model: claude-opus-4-7
-model_fallback: ["claude-opus-4-6", "claude-sonnet-4-6", "gpt-5", "gpt-4.5"]
 color: blue
 tools: ["Read", "Glob", "Grep", "Edit", "Write", "sdd_write_design", "sdd_generate_all_diagrams", "sdd_generate_diagram"]
 ---

package/.apm/agents/implementer.agent.md

@@ -2,8 +2,6 @@
 name: implementer
 description: Use this agent to generate implementation plans, quality checklists, test stubs, and infrastructure scaffolding from specifications.
 
-model: claude-sonnet-4-6
-model_fallback: ["codex", "gpt-5", "claude-opus-4-6"]
 color: blue
 tools: ["Read", "Glob", "Grep", "Edit", "Write", "MultiEdit", "Bash", "sdd_implement", "sdd_checklist", "sdd_generate_tests", "sdd_generate_pbt", "sdd_generate_iac", "sdd_generate_dockerfile", "sdd_generate_devcontainer", "sdd_setup_local_env", "sdd_setup_codespaces"]
 ---

package/.apm/agents/quality-reviewer.agent.md

@@ -2,8 +2,6 @@
 name: quality-reviewer
 description: Phase 6 agent that runs completeness audit, cross-analysis, and compliance checks. Produces ANALYSIS.md with gate decision (APPROVE/CONDITIONAL/REJECT) and COMPLIANCE.md.
 
-model: claude-opus-4-7
-model_fallback: ["claude-opus-4-6", "claude-sonnet-4-6", "gpt-5", "gpt-4.5"]
 color: red
 tools: ["Read", "Glob", "Grep", "Edit", "Write", "Bash", "sdd_run_analysis", "sdd_cross_analyze", "sdd_compliance_check", "sdd_check_sync", "sdd_metrics"]
 ---

package/.apm/agents/release-engineer.agent.md

@@ -2,8 +2,6 @@
 name: release-engineer
 description: Use this agent to prepare features for release — run blocking gates, generate documentation, create PR, and export work items.
 
-model: claude-haiku-4-5
-model_fallback: ["claude-sonnet-4-6", "gpt-4.5"]
 color: green
 tools: ["Read", "Glob", "Grep", "Edit", "Write", "Bash", "sdd_create_pr", "sdd_generate_all_docs", "sdd_generate_docs", "sdd_generate_api_docs", "sdd_generate_runbook", "sdd_generate_onboarding", "sdd_export_work_items"]
 ---

package/.apm/agents/requirements-engineer.agent.md

@@ -2,10 +2,8 @@
 name: requirements-engineer
 description: Use this agent to analyze raw input and produce validated FRD and NFRD documents ready for sdd_init.
 
-model: claude-opus-4-7
-model_fallback: ["claude-opus-4-6", "claude-sonnet-4-6", "gpt-5", "gpt-4.5"]
 color: magenta
-tools: ["Read", "Glob", "Grep", "Write", "Edit", "sdd_discover", "sdd_import_document", "sdd_import_transcript", "sdd_batch_import", "sdd_validate_ears"]
+tools: ["Read", "Glob", "Grep", "Write", "Edit", "sdd_init", "sdd_discover", "sdd_import_document", "sdd_import_transcript", "sdd_batch_import", "sdd_validate_ears"]
 ---
 
 <example>

package/.apm/agents/research-analyst.agent.md

@@ -2,8 +2,6 @@
 name: research-analyst
 description: Use this agent to gather technical context before specification. Scans codebases, imports documents, and produces RESEARCH.md.
 
-model: claude-sonnet-4-6
-model_fallback: ["claude-opus-4-6", "gpt-5", "gpt-4.5"]
 color: cyan
 tools: ["Read", "Glob", "Grep", "Bash", "WebFetch", "WebSearch", "sdd_scan_codebase", "sdd_discover", "sdd_research", "sdd_import_document", "sdd_import_transcript", "sdd_check_ecosystem"]
 ---

package/.apm/agents/sdd-clarify.agent.md

@@ -2,8 +2,6 @@
 name: sdd-clarify
 description: Use this agent to find and resolve ambiguities in requirements, validate EARS patterns, and produce a clarification log.
 
-model: claude-opus-4-7
-model_fallback: ["claude-opus-4-6", "claude-sonnet-4-6", "gpt-5", "gpt-4.5"]
 color: yellow
 tools: ["Read", "Glob", "Grep", "Edit", "sdd_clarify", "sdd_validate_ears", "sdd_turnkey_spec"]
 ---

package/.apm/agents/sdd-init.agent.md

@@ -2,8 +2,6 @@
 name: sdd-init
 description: Use this agent to initialize the SDD pipeline for a new feature. Creates the .specs/ directory structure and CONSTITUTION.md.
 
-model: claude-haiku-4-5
-model_fallback: ["claude-sonnet-4-6", "gpt-4.5"]
 color: green
 tools: ["Read", "Glob", "Grep", "Bash", "sdd_init", "sdd_scan_codebase"]
 ---

package/.apm/agents/spec-engineer.agent.md

@@ -2,18 +2,16 @@
 name: spec-engineer
 description: Phase 2 agent that writes SPECIFICATION.md using EARS notation. Every requirement gets a unique REQ-ID, one of the 6 EARS patterns, and measurable acceptance criteria.
 
-model: claude-opus-4-7
-model_fallback: ["claude-opus-4-6", "claude-sonnet-4-6", "gpt-5", "gpt-4.5"]
 color: magenta
 tools: ["Read", "Glob", "Grep", "Edit", "Write", "sdd_write_spec", "sdd_turnkey_spec", "sdd_validate_ears", "sdd_figma_to_spec"]
 ---
 
 <example>
-Context: Research phase is complete, ready to specify
+Context: Discovery is complete, ready to specify
 user: "Write the specification for feature 001"
 assistant: "I'll create SPECIFICATION.md with EARS requirements and validate all patterns."
 <commentary>
-Post-research specification writing is Phase 2.
+Post-discovery specification writing is the Specify phase.
 </commentary>
 </example>
 

package/.apm/agents/specky-onboarding.agent.md

@@ -2,10 +2,8 @@
 name: specky-onboarding
 description: Default entry point and interactive wizard for Specky SDD. Detects project context, explains the plugin, guides through project type selection, work mode, and branch setup. Triggered when user says "specky" without specifics.
 
-model: claude-haiku-4-5
-model_fallback: ["claude-sonnet-4-6", "gpt-4.5"]
 color: green
-tools: ["Read", "Glob", "Grep", "Bash", "Write", "sdd_get_status", "sdd_scan_codebase", "sdd_check_ecosystem", "sdd_context_status"]
+tools: ["Read", "Glob", "Grep", "Bash", "Write", "sdd_get_status", "sdd_scan_codebase", "sdd_check_ecosystem", "sdd_context_status", "sdd_auto_pipeline", "sdd_batch_import", "sdd_create_branch", "sdd_discover", "sdd_figma_to_spec", "sdd_import_document", "sdd_import_transcript"]
 ---
 
 <example>
@@ -67,8 +65,8 @@ You are the Specky SDD onboarding wizard. You are the **default entry point** wh
 - 13 agents with purpose and phase
 - 22 prompts grouped by use case
 - 8 skills with trigger conditions
-- 57 MCP tools by category
-- 14 hook scripts (10 advisory + 4 validation)
+- 58 MCP tools by category
+- 16 hook scripts (12 advisory + 4 blocking)
 
 **Hard rules:**
 - Always detect existing context first (never assume empty project)

package/.apm/agents/specky-orchestrator.agent.md

@@ -2,8 +2,6 @@
 name: specky-orchestrator
 description: Master agent that coordinates the full SDD 10-phase pipeline end-to-end, routing to phase agents, validating artifacts between phases, and enforcing hooks and LGTM gates.
 
-model: claude-sonnet-4-6
-model_fallback: ["claude-opus-4-6", "gpt-5", "gpt-4.5"]
 color: purple
 tools: ["Read", "Glob", "Grep", "Bash", "Task", "sdd_get_status", "sdd_advance_phase", "sdd_checkpoint", "sdd_validate_ears", "sdd_model_routing", "sdd_context_status"]
 ---
@@ -60,4 +58,4 @@ You are the Specky SDD pipeline orchestrator. You coordinate the full 10-phase p
 - Always pause at LGTM gates (Phases 2, 4, 5) for human review
 - Always checkpoint before advancing to next phase
 - Never modify artifacts — you coordinate, not author
-- Report model routing hint per phase (Haiku/Sonnet/Opus)
+- Report model routing hint per phase using capability classes (fast, balanced, reasoning-focused). Let the user pick any available model.

package/.apm/agents/task-planner.agent.md

@@ -2,8 +2,6 @@
 name: task-planner
 description: Phase 5 agent that writes TASKS.md with dependency-resolved task sequences, REQ-ID traceability, complexity estimates, and parallel markers. Also generates CHECKLIST.md.
 
-model: claude-sonnet-4-6
-model_fallback: ["claude-opus-4-6", "gpt-5", "gpt-4.5"]
 color: orange
 tools: ["Read", "Glob", "Grep", "Edit", "Write", "sdd_write_tasks", "sdd_checklist"]
 ---

package/.apm/agents/test-verifier.agent.md

@@ -2,8 +2,6 @@
 name: test-verifier
 description: Use this agent to verify test coverage, detect phantom completions, and check spec-code drift.
 
-model: claude-opus-4-7
-model_fallback: ["claude-opus-4-6", "claude-sonnet-4-6", "gpt-5", "gpt-4.5"]
 color: yellow
 tools: ["Read", "Glob", "Grep", "Edit", "Write", "Bash", "sdd_verify_tests", "sdd_verify_tasks", "sdd_check_sync", "sdd_validate_ears", "sdd_get_status"]
 ---

package/.apm/hooks/scripts/branch-validator.sh

@@ -2,12 +2,15 @@
 # branch-validator.sh — Validate git branch matches expected pattern for current tool.
 #
 # Type: Mixed (see below) | Trigger: PreToolUse | All phases
-#   • Advisory (exit 0, warn) for sdd_* pipeline tools
-#   • BLOCKING (exit 2) for Write|Edit|MultiEdit when a pipeline is active
-#     on a non-spec branch — prevents the SIFAP-style bypass where user
-#     edits code from impl/* instead of spec/NNN-*.
+#   • Advisory (exit 0, warn) for sdd_* pipeline tools (always).
+#   • For native edit tools (Write|Edit|MultiEdit) on a wrong branch during
+#     an active pipeline:
+#       - default (SPECKY_GUARD unset OR "off"/"advisory") → warn, exit 0
+#       - SPECKY_GUARD=strict                              → exit 2 (BLOCK)
 #
-# Escape hatch: SPECKY_GUARD=off allows bypass with warning (deprecated v3.6).
+# Rationale (rc.13): rc.10-rc.12 defaulted Write|Edit|MultiEdit to BLOCKING,
+# which burned pilots who run `specky init` and then edit anything from
+# develop/main. Enforcement is now explicit opt-in.
 
 set -euo pipefail
 
@@ -15,51 +18,53 @@ TOOL="${SDD_TOOL_NAME:-unknown}"
 CURRENT_BRANCH=$(git rev-parse --abbrev-ref HEAD 2>/dev/null | tr -d '\n' || true)
 [ -z "$CURRENT_BRANCH" ] && CURRENT_BRANCH="unknown"
 
-# ── BLOCKING path for native edit tools when pipeline is active ──
+# Mode selection (rc.13): strict = block, anything else = advisory.
+SPECKY_GUARD_MODE="${SPECKY_GUARD:-advisory}"
+if [ "$SPECKY_GUARD_MODE" = "off" ]; then
+  SPECKY_GUARD_MODE="advisory"
+fi
+
+# ── Branch enforcement for native edit tools when pipeline is active ──
 if [[ "$TOOL" == "Write" || "$TOOL" == "Edit" || "$TOOL" == "MultiEdit" ]]; then
-  # Check if a pipeline is active (same logic as pipeline-guard)
   LATEST=$(ls -td .specs/*/ 2>/dev/null | head -1 || true)
   if [ -z "$LATEST" ]; then exit 0; fi
   STATE="$LATEST/.sdd-state.json"
   if [ ! -f "$STATE" ]; then exit 0; fi
 
-  # Escape hatch
-  if [ "${SPECKY_GUARD:-on}" = "off" ]; then
-    echo "⚠️  [branch-validator] SPECKY_GUARD=off — allowing edit on '$CURRENT_BRANCH'" >&2
-    exit 0
-  fi
-
   PHASE="?"
   if command -v jq >/dev/null 2>&1; then
     PHASE=$(jq -r '.phase // "?"' "$STATE" 2>/dev/null || echo "?")
   fi
 
+  FEATURE=$(basename "$LATEST")
+
+  emit_violation() {
+    local expected="$1"
+    if [ "$SPECKY_GUARD_MODE" = "strict" ]; then
+      echo "" >&2
+      echo "🚫 [branch-validator] BLOCKED — $TOOL on '$CURRENT_BRANCH' (phase $PHASE expects $expected) [SPECKY_GUARD=strict]" >&2
+      echo "   Feature:  $FEATURE" >&2
+      echo "   Fix:      git checkout $expected   (or: git checkout -b spec/$FEATURE)" >&2
+      echo "   Disable:  unset SPECKY_GUARD" >&2
+      exit 2
+    fi
+    echo "⚠️  [branch-validator] $TOOL on '$CURRENT_BRANCH' during phase $PHASE (expected $expected). Enforce with: export SPECKY_GUARD=strict" >&2
+  }
+
   case "$PHASE" in
     0|1|2|3|4|5|6|7)
       if [[ "$CURRENT_BRANCH" != spec/* ]]; then
-        echo "" >&2
-        echo "🚫 [branch-validator] BLOCKED — edit on non-spec branch during active pipeline" >&2
-        echo "" >&2
-        echo "   Branch:   $CURRENT_BRANCH" >&2
-        echo "   Phase:    $PHASE (expects spec/NNN-*)" >&2
-        echo "   Feature:  $(basename "$LATEST")" >&2
-        echo "" >&2
-        echo "   Fix:      git checkout -b spec/$(basename "$LATEST")" >&2
-        echo "   Bypass:   export SPECKY_GUARD=off  (deprecated, removed v3.6)" >&2
-        echo "" >&2
-        exit 2
+        emit_violation "spec/$FEATURE"
       fi
       ;;
     8)
       if [[ "$CURRENT_BRANCH" != "develop" ]]; then
-        echo "🚫 [branch-validator] BLOCKED — Phase 8 requires 'develop' branch (you are on '$CURRENT_BRANCH')" >&2
-        exit 2
+        emit_violation "develop"
       fi
       ;;
     9)
       if [[ "$CURRENT_BRANCH" != "stage" ]]; then
-        echo "🚫 [branch-validator] BLOCKED — Phase 9 requires 'stage' branch (you are on '$CURRENT_BRANCH')" >&2
-        exit 2
+        emit_violation "stage"
       fi
       ;;
   esac

package/.apm/hooks/scripts/phase-gate.sh

@@ -56,7 +56,7 @@ case "$TOOL" in
   sdd_write_tasks)
     check_created "TASKS.md" "TASKS.md"
     check_has_content "TASKS.md" "REQ-[A-Z]+-[0-9]+" "REQ-ID traceability" 1
-    check_has_content "TASKS.md" "TASK-[0-9]+" "Task IDs" 3
+    check_has_content "TASKS.md" "T-?[0-9]{3}" "Task IDs" 3
     ;;
   sdd_verify_tests|sdd_verify_tasks)
     check_created "VERIFICATION.md" "VERIFICATION.md"

package/.apm/hooks/scripts/pipeline-guard.sh

@@ -1,14 +1,19 @@
 #!/bin/bash
-# pipeline-guard.sh — Enforce Rule #8: all work flows through @specky-orchestrator.
+# pipeline-guard.sh — Remind about Rule #8: all work flows through @specky-orchestrator.
 # Target: Claude Code (.claude/hooks/) + GitHub Copilot (.github/hooks/specky/)
-# Type: BLOCKING (exit 2) | Trigger: UserPromptSubmit
+# Type: ADVISORY by default (warn, exit 0) | Trigger: UserPromptSubmit
 # Phase: any — runs on every user prompt when .specs/ has an active pipeline
 #
-# Behavior:
-#   • If no .specs/ OR no active .sdd-state.json → exit 0 (pipeline-guard OFF).
-#   • If SPECKY_GUARD=off env var is set → exit 0 with warning.
-#   • If user prompt mentions orchestrator / onboarding / specky → exit 0.
-#   • If user prompt looks like a free-form code/build/edit request → exit 2 (BLOCK).
+# Behavior (rc.13 — flipped polarity after field incident):
+#   • If no .specs/ OR no active .sdd-state.json → exit 0 silently.
+#   • Default (SPECKY_GUARD unset OR "off"/"advisory") → print warning, exit 0.
+#   • Only when SPECKY_GUARD=strict → exit 2 (BLOCK) on free-form edit prompts.
+#
+# Rationale: rc.10-rc.12 defaulted this to BLOCKING, which fired on every
+# developer prompt containing common words (create, add, fix, test, install…).
+# Copilot surfaced every block as "Blocked by Pre-Tool Use hook", making the
+# IDE unusable. Enforcement is now explicit opt-in until the UX for
+# resuming/routing through @specky-orchestrator is polished.
 #
 # The user's prompt is read from stdin (Claude Code passes prompt as JSON on stdin)
 # OR from env var $CLAUDE_USER_PROMPT (Copilot). We try both.
@@ -18,12 +23,34 @@
 
 set -euo pipefail
 
-# ── Escape hatch ─────────────────────────────────────────
-if [ "${SPECKY_GUARD:-on}" = "off" ]; then
-  echo "⚠️  [pipeline-guard] SPECKY_GUARD=off — bypass allowed (deprecated, removed in v3.6)" >&2
+# ── rc.14: Copilot compatibility guard ───────────────────
+# VS Code Copilot reads .claude/settings.json hooks and treats SessionStart/
+# UserPromptSubmit as PreToolUse. When that happens, this script is invoked for
+# every tool call (Read, Glob, Grep…). The script tries to read the user prompt
+# from stdin but Copilot sends tool-call data instead → jq fails → cat hangs
+# → 5s timeout → "Blocked by Pre-Tool Use hook".
+#
+# Detection: Claude Code sets CLAUDE_PROJECT_DIR for its hooks. Copilot doesn't.
+# Also: if SDD_TOOL_NAME is set, we're being called as PreToolUse (not our event).
+if [ -z "${CLAUDE_PROJECT_DIR:-}" ] && [ -n "${SDD_TOOL_NAME:-}" ]; then
+  exit 0
+fi
+# Fallback: if no stdin is available and no CLAUDE_USER_PROMPT, skip gracefully.
+# Non-interactive stdin may contain a real prompt JSON payload in Claude Code
+# or in the test harness, so only skip when stdin is interactive.
+if [ -z "${CLAUDE_USER_PROMPT:-}" ] && [ -z "${CLAUDE_PROJECT_DIR:-}" ] && [ -t 0 ]; then
   exit 0
 fi
 
+# ── Mode selection ───────────────────────────────────────
+# strict  → block free-form edit prompts (exit 2)
+# advisory|off|<unset> → warn only (exit 0)
+SPECKY_GUARD_MODE="${SPECKY_GUARD:-advisory}"
+if [ "$SPECKY_GUARD_MODE" = "off" ]; then
+  # Back-compat alias: treat explicit "off" as advisory (still warn).
+  SPECKY_GUARD_MODE="advisory"
+fi
+
 # ── Check if pipeline is active ──────────────────────────
 LATEST=$(ls -td .specs/*/ 2>/dev/null | head -1 || true)
 if [ -z "$LATEST" ]; then
@@ -72,29 +99,35 @@ if echo "$PROMPT_LC" | grep -qE '^(what|why|how|where|when|show|list|explain|des
   exit 0
 fi
 
-# ── Blocklist: free-form code / build / edit requests ──
-# Keywords that trigger when a pipeline is active.
+# ── Edit-intent detection ──
+# Keywords that hint at free-form code/build/edit requests when a pipeline is active.
 # Pattern: keyword must be preceded and followed by a non-alphanumeric (or bounds).
 if echo "$PROMPT_LC" | grep -qE '(^|[^a-z0-9])(implement|create|build|write|code|fix|add|refactor|deploy|release|merge|commit|push|test|install|setup|configure)([^a-z0-9]|$)'; then
-  echo "" >&2
-  echo "🚫 [pipeline-guard] BLOCKED — active Specky pipeline detected" >&2
-  echo "" >&2
-  echo "   Feature:    $FEATURE" >&2
-  echo "   Phase:      $PHASE" >&2
-  echo "   State:      $STATE" >&2
-  echo "" >&2
-  echo "   Rule #8 (copilot-instructions.md): ALL work must flow through @specky-orchestrator." >&2
-  echo "" >&2
-  echo "   What to do:" >&2
-  echo "     • Resume:   invoke @specky-orchestrator (Copilot) or /specky-orchestrate (Claude)" >&2
-  echo "     • Status:   npx specky status" >&2
-  echo "     • Help:     invoke @specky-onboarding (Copilot) or /specky-onboarding (Claude)" >&2
-  echo "" >&2
-  echo "   Emergency bypass (logs warning, deprecated in v3.6):" >&2
-  echo "     export SPECKY_GUARD=off" >&2
-  echo "" >&2
-  exit 2
+  if [ "$SPECKY_GUARD_MODE" = "strict" ]; then
+    echo "" >&2
+    echo "🚫 [pipeline-guard] BLOCKED — active Specky pipeline detected (SPECKY_GUARD=strict)" >&2
+    echo "" >&2
+    echo "   Feature:    $FEATURE" >&2
+    echo "   Phase:      $PHASE" >&2
+    echo "   State:      $STATE" >&2
+    echo "" >&2
+    echo "   Rule #8 (copilot-instructions.md): ALL work must flow through @specky-orchestrator." >&2
+    echo "" >&2
+    echo "   What to do:" >&2
+    echo "     • Resume:   invoke @specky-orchestrator (Copilot) or /specky-orchestrate (Claude)" >&2
+    echo "     • Status:   specky status   (or: npx specky status)" >&2
+    echo "     • Help:     invoke @specky-onboarding (Copilot) or /specky-onboarding (Claude)" >&2
+    echo "" >&2
+    echo "   Disable enforcement (default):" >&2
+    echo "     unset SPECKY_GUARD      # or: export SPECKY_GUARD=advisory" >&2
+    echo "" >&2
+    exit 2
+  fi
+
+  # Advisory mode (default): warn but allow.
+  echo "⚠️  [pipeline-guard] active Specky pipeline ($FEATURE, phase $PHASE) — consider routing through @specky-orchestrator." >&2
+  echo "   Enforce blocking with: export SPECKY_GUARD=strict" >&2
 fi
 
-# Default: allow (ambiguous prompt — err on side of not blocking legitimate work)
+# Default: allow.
 exit 0

package/.apm/hooks/scripts/session-banner.sh

@@ -10,6 +10,14 @@
 
 set -euo pipefail
 
+# ── rc.14: Copilot compatibility guard ───────────────────
+# VS Code Copilot reads .claude/settings.json hooks and treats SessionStart as
+# PreToolUse. This script is advisory-only but Copilot surfaces any output from
+# a hook as a block warning. Skip entirely when not in Claude Code context.
+if [ -z "${CLAUDE_PROJECT_DIR:-}" ]; then
+  exit 0
+fi
+
 SPECS_DIR=".specs"
 [ -d "$SPECS_DIR" ] || exit 0
 
@@ -43,7 +51,7 @@ printf "│ Branch:   %-50s│\n" "$BRANCH"
 echo "├─────────────────────────────────────────────────────────────┤"
 echo "│ Resume:   @specky-orchestrator (Copilot)                     │"
 echo "│           /specky-orchestrate (Claude Code)                  │"
-echo "│ Status:   npx specky status                                  │"
+echo "│ Status:   specky status  (or npx specky status)             │"
 echo "│ New work: @specky-onboarding                                 │"
 echo "╰─────────────────────────────────────────────────────────────╯"
 echo ""

package/.apm/instructions/copilot-instructions.instructions.md

@@ -6,7 +6,7 @@ This project uses Spec-Driven Development (SDD) via the Specky pipeline.
 
 1. **EARS notation is mandatory.** Every requirement must follow one of the 6 EARS patterns.
 2. **REQ-ID traceability is non-negotiable.** Every test, task, and design decision traces to a REQ-ID.
-3. **Model routing matters.** Use Haiku for scaffolding (Phase 0, 9), Sonnet for iteration (Phase 1, 5-7), Opus for reasoning (Phase 2-4, 8).
+3. **Model routing matters.** Use a fast model class for lightweight scaffolding (Phase 0, 9), a balanced model class for iterative delivery (Phase 1, 5-7), and a reasoning-focused model class for complex analysis and design (Phase 2-4, 8). The user chooses any available model.
 4. **Never skip hooks.** Blocking hooks (security-scan, release-gate) must pass before release.
 5. **Artifacts live in `.specs/NNN-feature/`.** CONSTITUTION.md, RESEARCH.md, SPECIFICATION.md, DESIGN.md, TASKS.md, VERIFICATION.md, ANALYSIS.md.
 6. **One branch per spec.** Create `spec/NNN-feature-name` from `develop` for all pipeline work (Phases 0-7). All `.specs/` artifacts are created on this branch. Merge to `develop` after verification, then `stage` for QA and release gates, then `main` for production. Never commit spec work directly to develop, stage, or main.
@@ -61,7 +61,7 @@ Hooks fire automatically on MCP tool calls:
 
 ## MCP Server
 
-The specky-sdd MCP server (57 tools) is configured in .vscode/mcp.json and runs via npx.
+The specky-sdd MCP server (58 tools) is configured in .vscode/mcp.json and runs via npx.
 
 ## EARS Patterns
 

package/.apm/skills/sdd-markdown-standard/SKILL.md

@@ -193,7 +193,7 @@ version: "1.2.3"
 revision_history:
   - version: "1.2.3"
     date: "2026-04-15"
-    author: "Claude Opus"
+    author: "GitHub Copilot"
     changes: "Added security requirements REQ-SEC-001 to REQ-SEC-005"
   - version: "1.2.2"
     date: "2026-04-14"
@@ -201,7 +201,7 @@ revision_history:
     changes: "Clarified API response format in REQ-API-003"
   - version: "1.2.1"
     date: "2026-04-13"
-    author: "Claude Opus"
+    author: "GitHub Copilot"
     changes: "Fixed typo in REQ-AUTH-002"
 ```
 

package/.apm/skills/sdd-pipeline/SKILL.md

@@ -51,32 +51,32 @@ EARS (Easy Approach to Requirements Syntax) provides five core patterns plus a c
 6. **Complex** — Combination patterns for intricate business logic.
    - Example: "If user role is editor, when publish button is clicked, then system shall validate content and if valid shall queue for approval."
 
-## Model Routing Table
+## Model Recommendation Table
 
-Route specification and implementation tasks to models based on phase complexity:
+Recommend a model capability class per phase. Do not hardcode a vendor-specific model ID. Let the user choose any available model.
 
-| Phase | Model | Reasoning |
-|-------|-------|-----------|
-| 0 (Init) | Haiku | Basic scope definition, lightweight |
-| 1 (Discover) | Sonnet | Multi-source synthesis, ecosystem analysis |
-| 2 (Specify) | Opus | Complex requirement formalization, EARS patterns |
-| 3 (Clarify) | Opus | Interactive refinement, stakeholder context |
-| 4 (Design) | Opus | Architecture decisions, multi-component systems |
-| 5 (Tasks) | Sonnet | Work breakdown, dependency mapping |
-| 6 (Analyze) | Sonnet | Cross-artifact analysis, compliance checks |
-| 7 (Implement) | Sonnet | Code scaffolding, quality checklists |
-| 8 (Verify) | Opus | Coverage analysis, drift detection |
-| 9 (Release) | Haiku | Final gates, documentation assembly |
+| Phase | Recommended Class | Reasoning |
+|-------|-------------------|-----------|
+| 0 (Init) | Fast | Basic scope definition and scaffolding |
+| 1 (Discover) | Balanced | Multi-source synthesis and ecosystem analysis |
+| 2 (Specify) | Reasoning-focused | Complex requirement formalization and EARS patterns |
+| 3 (Clarify) | Reasoning-focused | Ambiguity detection and refinement |
+| 4 (Design) | Reasoning-focused | Architecture decisions and trade-off analysis |
+| 5 (Tasks) | Balanced | Work breakdown and dependency mapping |
+| 6 (Analyze) | Balanced | Cross-artifact analysis and compliance checks |
+| 7 (Implement) | Balanced | Iterative implementation with executable feedback |
+| 8 (Verify) | Reasoning-focused | Coverage analysis and drift detection |
+| 9 (Release) | Fast | Deterministic final gates and documentation assembly |
 
-## Extended Thinking Impact
+## Extended Reasoning Guidance
 
 Reference: **arXiv:2502.08235** — "Extended Thinking and Specification Quality in Large Language Models"
 
-Key finding for Phase 7 (Implementation): Enabling extended thinking (chain-of-thought) reduces quality by 30% while increasing cost by 43% in code generation tasks. Recommendation: Use standard inference for Phase 7 scaffolding; reserve extended thinking for Phase 8 verification.
+For Phase 7 (Implementation), prioritize fast feedback loops over long reasoning traces. Reserve deeper reasoning for ambiguous design and verification tasks.
 
 ## Hook System
 
-The pipeline includes 14 automation hooks for customization:
+The pipeline includes 16 automation hooks for customization:
 
 **Blocking Hooks** (workflow stops if hook fails — exit code 2):
 1. `artifact-validator` — Pre-tool: blocks if required artifacts missing
@@ -126,11 +126,14 @@ Hooks are configured in `sdd-hooks.json` with PreToolUse and PostToolUse matcher
 
 **Agent-based:**
 ```
-@specky-sdd-agent init feature-name
-@specky-sdd-agent specify --input=RESEARCH.md
-@specky-sdd-agent implement --model=sonnet
+@specky-orchestrator           (end-to-end pipeline coordinator)
+@sdd-init feature-name         (Phase 0 — initialize feature)
+@spec-engineer                 (Phase 2 — write SPECIFICATION.md)
+@implementer                   (Phase 7 — scaffold code + tests)
 ```
 
+See [copilot-instructions.instructions.md](../../instructions/copilot-instructions.instructions.md) for the full agent catalog.
+
 ## Workflow Entry Points
 
 - **Greenfield** — Start at Phase 0 with new project initialization
@@ -221,17 +224,17 @@ git push origin main --tags
 
 ## Agent Routing
 
-| Phase | Agent | Model |
-|-------|-------|-------|
-| Pre | @specky-onboarding | Haiku |
-| All | @specky-orchestrator | Sonnet |
-| 0 | @sdd-init | Haiku |
-| 1 | @research-analyst | Sonnet |
-| 2 | @spec-engineer | Opus |
-| 3 | @sdd-clarify | Opus |
-| 4 | @design-architect | Opus |
-| 5 | @task-planner | Sonnet |
-| 6 | @quality-reviewer | Sonnet |
-| 7 | @implementer | Sonnet |
-| 8 | @test-verifier | Opus |
-| 9 | @release-engineer | Haiku |
+| Phase | Agent | Recommended Class |
+|-------|-------|-------------------|
+| Pre | @specky-onboarding | Fast |
+| All | @specky-orchestrator | Balanced |
+| 0 | @sdd-init | Fast |
+| 1 | @research-analyst | Balanced |
+| 2 | @spec-engineer | Reasoning-focused |
+| 3 | @sdd-clarify | Reasoning-focused |
+| 4 | @design-architect | Reasoning-focused |
+| 5 | @task-planner | Balanced |
+| 6 | @quality-reviewer | Balanced |
+| 7 | @implementer | Balanced |
+| 8 | @test-verifier | Reasoning-focused |
+| 9 | @release-engineer | Fast |