@softspark/ai-toolkit 2.7.0 → 2.7.2

package/CHANGELOG.md

@@ -7,6 +7,28 @@ Versioning follows [Semantic Versioning](https://semver.org/).
 
 ---
 
+## v2.7.2 — Doc & Model-ID Consistency (2026-04-17)
+
+### Changed
+- **Centralized Claude model IDs** — introduced `DEFAULT_CLAUDE_MODELS` dict in `scripts/_common.py` (single source of truth: `claude-opus-4-7`, `claude-sonnet-4-6`, `claude-haiku-4-5`). `scripts/generate_aider_conf.py` now imports the constant instead of hardcoding `claude-sonnet-4-6`. Next Anthropic model bump touches one file. Agent frontmatter aliases (`opus`/`sonnet`/`haiku`) stay version-free and are resolved at runtime.
+- **`app/agents/ai-engineer.md`** now points to the `model-routing-patterns` skill as the single source of truth for current Claude IDs, cost tiers, and fallback chains instead of duplicating a tier table that would drift with every Anthropic release.
+
+### Fixed
+- **`GEMINI.md` missed 5 skills from v2.7.0** — `generate:all` was not re-run before the v2.7.0 release tag, so `GEMINI.md` did not list `prompt-caching-patterns`, `json-mode-patterns`, `content-moderation-patterns`, `model-routing-patterns`, or `/mcp-builder`. Regenerated.
+- **`manifest.json` stale skill counts** — `components.skills.description` and `modules.skills.description` still reported `94 skills (31 task + 31 hybrid + 32 knowledge)` instead of the current `99 skills (32 task + 31 hybrid + 36 knowledge)`. Fixed in 2 places.
+
+---
+
+## v2.7.1 — Eject Output-Styles Fix (2026-04-17)
+
+### Fixed
+- **`ai-toolkit eject` now exports `app/output-styles/*.md`** — pre-existing gap dating back to when output-styles were introduced. `scripts/eject.py` copied agents, skills, rules, constitution, and ARCHITECTURE, but silently skipped the `output-styles/` directory. Standalone `.claude/` produced by eject had no system-prompt styles (no `golden-rules`, and after v2.7.0 no `learning`/`explanatory`). Now eject writes `.claude/output-styles/` with all source `.md` files and includes the count in the summary line: `Ejected: N agents, N skills, N rules, N output style(s)`.
+
+### Added
+- **2 `tests/test_eject.bats` cases** — `eject copies output-styles directory with all source styles` (also asserts `golden-rules.md` baseline) and `eject reports output-style count in summary line`. Test count: 664 → 666.
+
+---
+
 ## v2.7.0 — Anthropic Ecosystem Alignment (2026-04-17)
 
 ### Added

package/README.md

@@ -6,17 +6,15 @@
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
 [![Skills](https://img.shields.io/badge/skills-99-brightgreen)](app/skills/)
 [![Agents](https://img.shields.io/badge/agents-44-blue)](app/agents/)
-[![Tests](https://img.shields.io/badge/tests-664%20passing-success)](tests/)
+[![Tests](https://img.shields.io/badge/tests-666%20passing-success)](tests/)
 
 ---
 
-## What's New in v2.7.0
+## What's New in v2.7.2
 
-- **5 new skills aligned with Anthropic ecosystem** — `/mcp-builder` (task skill: 4-phase MCP server build methodology, research → implement → test → evaluate) and 4 knowledge skills auto-loaded for Claude API work: `prompt-caching-patterns`, `json-mode-patterns`, `content-moderation-patterns`, `model-routing-patterns`. Skill count: 94 → 99.
-- **2 new output styles** in `app/output-styles/` — `learning` (interactive "your turn" prompts on meaningful decisions) and `explanatory` (★ Insight blocks surfacing implementation trade-offs). Both stack with the existing `golden-rules` style.
-- **`/review` secrets scanner hardened** — `diff-analyzer.py` now tracks actual file line numbers (not diff offsets) per hunk, detects renames via `git diff --numstat -z`, surfaces a `warnings[]` array when falling back from branch diff to staged, and expands secret patterns (JWT, PEM private keys, Google API, Slack tokens, GitHub fine-grained PAT, unquoted env-style). Category regex tightened with word boundaries and anchored path patterns (`docs/role-permissions.md` is no longer miscategorised as security). 4 fixture tests in `tests/test_review_diff_analyzer.bats`.
-- **Frontmatter Schema reference** — new section in `app/ARCHITECTURE.md` and canonical spec mirror at `kb/reference/agent-skills-spec.md` (in rag-mcp KB). Documents ai-toolkit's extensions beyond `agentskills.io/specification` and the comma-separated `allowed-tools` convention enforced by `codex_skill_adapter.py` and `audit_skills.py`.
-- **Test count:** 660 → 664.
+- **Centralized Claude model IDs** in `scripts/_common.py` (`DEFAULT_CLAUDE_MODELS` dict). Next Anthropic model bump touches one file instead of hunting through generators.
+- **`GEMINI.md` regenerated** to include 5 skills missed at v2.7.0 tag: `prompt-caching-patterns`, `json-mode-patterns`, `content-moderation-patterns`, `model-routing-patterns`, `/mcp-builder`.
+- **`manifest.json` skill count corrected** — `94 skills (31/31/32)` → `99 skills (32/31/36)` in both component and module descriptions.
 
 See [CHANGELOG.md](CHANGELOG.md) for full history.
 
@@ -144,7 +142,7 @@ ai-toolkit/
 │   └── ARCHITECTURE.md  # Full system design
 ├── kb/                  # Reference docs, procedures, plans
 ├── scripts/             # Validation, install, evaluation scripts
-├── tests/               # Bats test suite (664 tests)
+├── tests/               # Bats test suite (666 tests)
 └── CHANGELOG.md
 ```
 

package/app/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "ai-toolkit",
   "description": "Professional-grade Claude Code toolkit with persona presets, skill security auditor, expanded lifecycle hooks, experimental opt-in plugin packs, benchmark harvesting, and multi-tool support.",
-  "version": "2.7.0",
+  "version": "2.7.2",
   "author": {
     "name": "SoftSpark",
     "url": "https://github.com/softspark"

package/app/agents/ai-engineer.md

@@ -64,6 +64,8 @@ AI/ML integration specialist for production systems, including RAG pipeline desi
 | Complex reasoning | Large | Claude Opus, GPT-4 |
 | Local/private | Open | Llama, Mistral |
 
+For current Claude model IDs, cost tiers, and fallback chains see the `model-routing-patterns` skill — the single source of truth that gets bumped with each Anthropic release.
+
 ### Embedding Selection
 | Use Case | Model |
 |----------|-------|

package/llms-full.txt

@@ -4647,9 +4647,9 @@ Three tiers determine how to approach a task:
 
 | Type | Field | Invocation | Count |
 |------|-------|-----------|-------|
-| Task | `disable-model-invocation: true` | User via `/skill` only | 31 |
+| Task | `disable-model-invocation: true` | User via `/skill` only | 32 |
 | Hybrid | (neither) | User via `/skill` + agent knowledge | 31 |
-| Knowledge | `user-invocable: false` | Claude auto-loads | 32 |
+| Knowledge | `user-invocable: false` | Claude auto-loads | 36 |
 
 ## Multi-Agent Execution
 

package/manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "2.7.0",
+  "version": "2.7.2",
   "components": {
     "agents": {
       "description": "44 specialized agents (orchestrator, backend, frontend, security, devops, etc.)",
@@ -9,7 +9,7 @@
       "tags": ["core", "agents"]
     },
     "skills": {
-      "description": "94 skills (31 task + 31 hybrid + 32 knowledge)",
+      "description": "99 skills (32 task + 31 hybrid + 36 knowledge)",
       "path": "app/skills",
       "target": ".claude/skills",
       "type": "symlink",
@@ -90,7 +90,7 @@
       "default": true
     },
     "skills": {
-      "description": "94 skills (task, hybrid, knowledge)",
+      "description": "99 skills (task, hybrid, knowledge)",
       "default": true
     },
     "rules-common": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@softspark/ai-toolkit",
-  "version": "2.7.0",
+  "version": "2.7.2",
   "description": "Professional-grade AI coding toolkit: 99 skills, 44 agents, multi-platform support (Claude, Cursor, Windsurf, Copilot, Gemini, Cline, Roo Code, Aider, Augment, Google Antigravity, Codex CLI, opencode), machine-enforced safety constitution, persona presets, skill security auditor, expanded lifecycle hooks, 11 plugin packs, and benchmark tooling.",
   "keywords": [
     "claude",

package/scripts/_common.py

@@ -72,6 +72,19 @@ agents_dir: Path = app_dir / "agents"
 skills_dir: Path = app_dir / "skills"
 
 
+# ---------------------------------------------------------------------------
+# Default Claude model IDs (single source of truth — bump on Anthropic release)
+# Consumed by generators that need to emit a concrete model string.
+# Aliases like "opus"/"sonnet"/"haiku" in agent frontmatter are resolved by
+# the client at runtime and do NOT need updating here.
+# ---------------------------------------------------------------------------
+DEFAULT_CLAUDE_MODELS: dict[str, str] = {
+    "opus": "claude-opus-4-7",
+    "sonnet": "claude-sonnet-4-6",
+    "haiku": "claude-haiku-4-5",
+}
+
+
 # ---------------------------------------------------------------------------
 # Component filtering (for install/update)
 # ---------------------------------------------------------------------------

package/scripts/eject.py

@@ -7,6 +7,7 @@ target directory with no symlinks and no dependency on ai-toolkit.
 What it exports:
   - .claude/agents/*.md (real files, not symlinks)
   - .claude/skills/*/   (real directories, not symlinks)
+  - .claude/output-styles/*.md (system-prompt styles)
   - .claude/CLAUDE.md   (inlined rules)
   - .claude/constitution.md (full content)
   - .claude/ARCHITECTURE.md (full content)
@@ -90,6 +91,21 @@ def main() -> None:
             rule_count += 1
     print(f"  Inlined: {rule_count} rules into CLAUDE.md")
 
+    # -- Output styles: copy as real files -----------------------------------
+    print("## Output Styles")
+    style_count = 0
+    styles_src = app_dir / "output-styles"
+    if styles_src.is_dir():
+        styles_out = claude_dir / "output-styles"
+        styles_out.mkdir(parents=True, exist_ok=True)
+        for style in sorted(styles_src.glob("*.md")):
+            dest = styles_out / style.name
+            if dest.is_symlink():
+                dest.unlink()
+            shutil.copy2(style, dest)
+            style_count += 1
+    print(f"  Copied: {style_count} output style(s)")
+
     # -- Constitution and Architecture ---------------------------------------
     print("## Config Files")
     for filename in ("constitution.md", "ARCHITECTURE.md"):
@@ -104,7 +120,7 @@ def main() -> None:
     # -- Summary -------------------------------------------------------------
     print()
     print("========================")
-    print(f"Ejected: {agent_count} agents, {skill_count} skills, {rule_count} rules")
+    print(f"Ejected: {agent_count} agents, {skill_count} skills, {rule_count} rules, {style_count} output style(s)")
     print()
     print(f"The toolkit is now standalone in {target_dir.resolve()}/.claude/")
     print("You can safely run: npm uninstall -g @softspark/ai-toolkit")

package/scripts/generate_aider_conf.py

@@ -10,8 +10,11 @@ from pathlib import Path
 
 sys.path.insert(0, str(Path(__file__).resolve().parent))
 
+from _common import DEFAULT_CLAUDE_MODELS
+
 
 def main() -> None:
+    default_model = DEFAULT_CLAUDE_MODELS["sonnet"]
     print("# Aider configuration generated by ai-toolkit")
     print("# Regenerate using: ai-toolkit install --local")
     print()
@@ -34,8 +37,8 @@ def main() -> None:
     print('  - "go: go vet"')
     print()
     print("# Model configuration")
-    print("model: claude-sonnet-4-6")
-    print("editor-model: claude-sonnet-4-6")
+    print(f"model: {default_model}")
+    print(f"editor-model: {default_model}")
     print()
     print("# Auto-commit message formatting (matches our conventional commits)")
     print(