Bump version to 0.1.2

Release highlights:
- Exit-127 detection prevents false test penalties in Copeland scoring
- 4 example projects (A* Python/TS, ML regression, ML classification)
- Scoring evaluation: n=73 runs, Cochran's Q p<0.0001
- Two-phase ensemble test generation workflow documented
- GitHub Packages publishing

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: unknown

CHANGELOG.md

@@ -5,10 +5,33 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-## [Unreleased]
+## [0.1.2] — 2026-03-30
 
 ### Added
-- `thinktank run` command — spawn N parallel Claude Code agents in isolated worktrees
+- Example projects: A* pathfinding (Python + TypeScript), ML regression, ML classification
+- Ensemble-generated test suite demonstrating two-phase workflow
+- Scoring evaluation: n=73 usable runs with Cochran's Q, Wilcoxon, Cliff's delta
+- README: documented two-phase ensemble test generation workflow
+- GitHub Packages publishing alongside npmjs.org
+
+### Fixed
+- Exit-127 (command not found) in test runner now marked as `skipped` instead of `failed` — prevents false penalties in Copeland scoring
+- Preflight check hard-errors on exit 127 instead of just warning — saves API tokens
+- Copeland scoring skips tests criterion when test command couldn't execute
+- `test_maze` assertion in A* examples (correct path length is 9, not 13)
+- CLI version read from package.json at runtime (no more stale hardcoded version)
+- Log messages: "Setting up agent environments" (was "Creating worktrees")
+
+## [0.1.1] — 2026-03-29
+
+### Fixed
+- Version bump for npm publish (0.1.0 was already claimed)
+- NPM trusted publishing requires NPM_TOKEN alongside OIDC
+
+## [0.1.0] — 2026-03-29
+
+### Added
+- `thinktank run` command — spawn N parallel Claude Code agents in isolated git clones
 - `thinktank apply` command — apply recommended or selected agent's changes
 - `thinktank apply --preview` — review diff with syntax highlighting before applying
 - `thinktank compare <A> <B>` — compare two agents' diffs side by side

package-lock.json

@@ -1,6 +1,6 @@
 {
   "name": "thinktank-ai",
-  "version": "0.1.1",
+  "version": "0.1.2",
   "description": "Ensemble AI coding — run N parallel Claude Code agents, select the best result via Copeland pairwise scoring",
   "type": "module",
   "bin": {