@plaited/agent-eval-harness 0.6.0 → 0.6.2

package/README.md

@@ -25,7 +25,7 @@ export ANTHROPIC_API_KEY=sk-...   # For Claude
 export GEMINI_API_KEY=...         # For Gemini
 ```
 
-Pre-built schemas are available in `.claude/skills/headless-adapters/schemas/` for Claude and Gemini.
+Pre-built schemas are available in `.plaited/skills/headless-adapters/schemas/` for Claude and Gemini.
 
 ### Core Commands
 
@@ -85,7 +85,7 @@ bunx @plaited/agent-eval-harness compare run1.jsonl run2.jsonl -o comparison.jso
 **Install skills** for use with AI coding agents:
 
 ```bash
-curl -fsSL https://raw.githubusercontent.com/plaited/skills-installer/main/install.sh | bash -s -- --agent <agent-name> --project agent-eval-harness
+curl -fsSL https://raw.githubusercontent.com/plaited/skills-installer/main/install.sh | bash -s -- --agents <agent-name> --project agent-eval-harness
 ```
 
 Replace `<agent-name>` with your agent: `claude`, `cursor`, `copilot`, `opencode`, `amp`, `goose`, `factory`
@@ -253,18 +253,20 @@ cat results.jsonl | your-scoring-script.ts
 ## Development
 
 ```bash
-bun install          # Install dependencies
-bun run check        # Type check + lint + format
-bun test             # Run unit tests
-
-# Run integration tests in Docker (requires API keys)
-ANTHROPIC_API_KEY=sk-... docker compose -f docker-compose.test.yml run --rm test
+bun install               # Install dependencies
+bun run check             # Type check + lint + format
+bun test                  # Run unit tests
+bun run test:integration  # Run integration tests (requires API keys)
+
+# Alternative: Run integration tests in Docker
+ANTHROPIC_API_KEY=sk-... GEMINI_API_KEY=... \
+  docker compose -f docker-compose.test.yml run --rm test
 ```
 
 ## Requirements
 
 - **Runtime:** Bun >= 1.2.9
-- **Schema:** JSON schema describing CLI agent interaction (see `.claude/skills/headless-adapters/schemas/`)
+- **Schema:** JSON schema describing CLI agent interaction (see `.plaited/skills/headless-adapters/schemas/`)
 - **API Key:** `ANTHROPIC_API_KEY` for Claude, `GEMINI_API_KEY` for Gemini
 
 ## License

package/bin/tests/cli.spec.ts

@@ -219,8 +219,8 @@ const SAMPLE_SUMMARY_JSONL = `{"id":"test-001","input":"Create a button","output
 {"id":"test-002","input":"Fix the bug","output":"I fixed the bug","toolCalls":["Read","Edit"],"duration":2567}
 {"id":"test-003","input":"Broken test","output":"","toolCalls":[],"duration":500}`
 
-const SAMPLE_CAPTURE_JSONL = `{"id":"test-001","input":"Create a button","output":"I created the button","trajectory":[{"type":"thought","content":"I'll create a button template","timestamp":100,"stepId":"test-001-step-1"},{"type":"tool_call","name":"Write","status":"completed","input":{"file_path":"src/button.tsx","content":"export const Button = () => <button>Click</button>"},"output":"File written","duration":234,"timestamp":150,"stepId":"test-001-step-2"},{"type":"message","content":"I created the button","timestamp":500,"stepId":"test-001-step-3"}],"metadata":{"category":"ui","agent":"claude-code-acp"},"timing":{"start":1704067200000,"end":1704067201234,"firstResponse":100},"toolErrors":false}
-{"id":"test-002","input":"Fix the bug","output":"I fixed the bug","trajectory":[{"type":"tool_call","name":"Read","status":"completed","input":{"file_path":"src/app.ts"},"output":"file contents...","duration":100,"timestamp":50,"stepId":"test-002-step-1"},{"type":"tool_call","name":"Edit","status":"completed","input":{"file_path":"src/app.ts","old_string":"bug","new_string":"fix"},"duration":150,"timestamp":200,"stepId":"test-002-step-2"},{"type":"message","content":"I fixed the bug","timestamp":400,"stepId":"test-002-step-3"}],"metadata":{"category":"bugfix","agent":"claude-code-acp"},"timing":{"start":1704067300000,"end":1704067302567},"toolErrors":false}`
+const SAMPLE_CAPTURE_JSONL = `{"id":"test-001","input":"Create a button","output":"I created the button","trajectory":[{"type":"thought","content":"I'll create a button template","timestamp":100,"stepId":"test-001-step-1"},{"type":"tool_call","name":"Write","status":"completed","input":{"file_path":"src/button.tsx","content":"export const Button = () => <button>Click</button>"},"output":"File written","duration":234,"timestamp":150,"stepId":"test-001-step-2"},{"type":"message","content":"I created the button","timestamp":500,"stepId":"test-001-step-3"}],"metadata":{"category":"ui","agent":"claude-headless"},"timing":{"start":1704067200000,"end":1704067201234,"firstResponse":100},"toolErrors":false}
+{"id":"test-002","input":"Fix the bug","output":"I fixed the bug","trajectory":[{"type":"tool_call","name":"Read","status":"completed","input":{"file_path":"src/app.ts"},"output":"file contents...","duration":100,"timestamp":50,"stepId":"test-002-step-1"},{"type":"tool_call","name":"Edit","status":"completed","input":{"file_path":"src/app.ts","old_string":"bug","new_string":"fix"},"duration":150,"timestamp":200,"stepId":"test-002-step-2"},{"type":"message","content":"I fixed the bug","timestamp":400,"stepId":"test-002-step-3"}],"metadata":{"category":"bugfix","agent":"claude-headless"},"timing":{"start":1704067300000,"end":1704067302567},"toolErrors":false}`
 
 // ============================================================================
 // Downstream Pattern Tests
@@ -429,7 +429,7 @@ describe('MCP server config parsing', () => {
   test('parses stdio MCP server config', () => {
     const json = '{"type":"stdio","name":"fs","command":"mcp-filesystem","args":["/data"],"env":[]}'
     const proc = Bun.spawn(
-      ['bun', CLI_PATH, 'capture', '/tmp/test.jsonl', 'bunx', 'claude-code-acp', '--mcp-server', json, '--help'],
+      ['bun', CLI_PATH, 'capture', '/tmp/test.jsonl', '--schema', './test-schema.json', '--mcp-server', json, '--help'],
       {
         stdout: 'pipe',
         stderr: 'pipe',
@@ -444,7 +444,7 @@ describe('MCP server config parsing', () => {
     const json =
       '{"type":"http","name":"api","url":"https://example.com/mcp","headers":[{"name":"Authorization","value":"Bearer token"}]}'
     const proc = Bun.spawn(
-      ['bun', CLI_PATH, 'capture', '/tmp/test.jsonl', 'bunx', 'claude-code-acp', '--mcp-server', json, '--help'],
+      ['bun', CLI_PATH, 'capture', '/tmp/test.jsonl', '--schema', './test-schema.json', '--mcp-server', json, '--help'],
       {
         stdout: 'pipe',
         stderr: 'pipe',
@@ -464,8 +464,8 @@ describe('MCP server config parsing', () => {
         CLI_PATH,
         'capture',
         '/tmp/test.jsonl',
-        'bunx',
-        'claude-code-acp',
+        '--schema',
+        './test-schema.json',
         '--mcp-server',
         json1,
         '--mcp-server',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@plaited/agent-eval-harness",
-  "version": "0.6.0",
+  "version": "0.6.2",
   "description": "CLI tool for capturing agent trajectories from headless CLI agents",
   "license": "ISC",
   "engines": {
@@ -54,11 +54,11 @@
     ]
   },
   "dependencies": {
-    "zod": "^4.3.5",
-    "@plaited/development-skills": "0.6.3"
+    "@plaited/development-skills": "0.6.5",
+    "zod": "^4.3.6"
   },
   "devDependencies": {
-    "@biomejs/biome": "2.3.11",
+    "@biomejs/biome": "2.3.12",
     "@types/bun": "1.3.6",
     "format-package": "7.0.0",
     "lint-staged": "16.2.7",