helixevo 0.2.6 → 0.2.8

package/README.md

@@ -12,7 +12,7 @@ HelixEvo builds on ideas from [EvoSkill](https://arxiv.org/abs/2603.02766) and [
 
 Every proposed change goes through:
 1. **3 independent LLM judges** (Task Completion, Correction Alignment, Side-Effect Check)
-2. **Regression testing** against golden cases
+2. **Regression testing** against skill tests
 3. **3-day canary deployment** with auto-rollback
 
 ## Prerequisites
@@ -57,7 +57,7 @@ npm link
 ## Quick Start
 
 ```bash
-# 1. Initialize — imports existing skills + generates golden cases
+# 1. Initialize — imports existing skills + generates skill tests
 helixevo init
 
 # 2. Capture failures from a session
@@ -80,7 +80,7 @@ helixevo dashboard
 | `helixevo watch` | Always-on learning: auto-capture + auto-evolve |
 | `helixevo metrics` | Correction rates, skill trends, evolution impact |
 | `helixevo health` | Network health: cohesion, coverage, balance, transfer |
-| `helixevo init` | Import existing skills + generate golden cases |
+| `helixevo init` | Import existing skills + generate skill tests |
 | `helixevo capture <session>` | Extract failures from a session file |
 | `helixevo evolve` | Evolve skills from captured failures |
 | `helixevo generalize` | Promote cross-project patterns ↑ |
@@ -126,7 +126,7 @@ All data is stored in `~/.helix/`:
 ├── failures.jsonl           # Captured failures
 ├── frontier.json            # Pareto frontier (top-k configurations)
 ├── evolution-history.json   # All evolution runs + proposals
-├── golden-cases.jsonl       # Regression test cases
+├── skill-tests.jsonl       # Regression test cases
 ├── skill-graph.json         # Cached network (nodes + edges)
 ├── canary-registry.json     # Active canary deployments
 ├── knowledge-buffer.json    # Research discoveries + drafts

package/dashboard/app/api/skills/route.ts

@@ -122,7 +122,7 @@ function analyzeNetworkAdaptation(
       })
       result.suggestions.push({
         type: 'rewire',
-        description: `Review golden cases for [${partners.join(', ')}] after this change`
+        description: `Review skill tests for [${partners.join(', ')}] after this change`
       })
     }
   }

package/dashboard/app/guide/page.tsx

@@ -1,6 +1,6 @@
 'use client'
 
-import { useState } from 'react'
+import { useState, useEffect, useRef } from 'react'
 
 // ─── Table of Contents ──────────────────────────────────────────
 const TOC = [
@@ -148,7 +148,7 @@ function ArchitectureDiagram() {
         <div className="guide-diagram-box guide-diagram-check" style={{ direction: 'ltr' }}>
           <div className="guide-diagram-box-label">Validate</div>
           <div className="guide-diagram-box-title">Regression Tests</div>
-          <div className="guide-diagram-box-desc">Golden cases + cross-skill</div>
+          <div className="guide-diagram-box-desc">Skill tests + cross-skill</div>
         </div>
         <div className="guide-diagram-arrow" style={{ direction: 'ltr' }}>←</div>
         <div className="guide-diagram-box guide-diagram-judge" style={{ direction: 'ltr' }}>
@@ -195,6 +195,30 @@ function HierarchyDiagram() {
 // ─── Main Guide Page ────────────────────────────────────────────
 export default function GuidePage() {
   const [activeSection, setActiveSection] = useState('overview')
+  const contentRef = useRef<HTMLDivElement>(null)
+
+  useEffect(() => {
+    const sectionIds = TOC.map(t => t.id)
+    const observer = new IntersectionObserver(
+      (entries) => {
+        // Find the most visible section
+        const visible = entries
+          .filter(e => e.isIntersecting)
+          .sort((a, b) => b.intersectionRatio - a.intersectionRatio)
+        if (visible.length > 0) {
+          setActiveSection(visible[0].target.id)
+        }
+      },
+      { rootMargin: '-80px 0px -60% 0px', threshold: [0, 0.25, 0.5] }
+    )
+
+    for (const id of sectionIds) {
+      const el = document.getElementById(id)
+      if (el) observer.observe(el)
+    }
+
+    return () => observer.disconnect()
+  }, [])
 
   return (
     <div className="guide-layout">
@@ -202,7 +226,7 @@ export default function GuidePage() {
       <nav className="guide-toc">
         <div className="guide-toc-header">
           <div className="guide-toc-title">Documentation</div>
-          <div className="guide-toc-version">v0.2.0</div>
+          <div className="guide-toc-version">v0.2.6</div>
         </div>
         {TOC.map(item => (
           <a
@@ -222,7 +246,7 @@ export default function GuidePage() {
         {/* Hero */}
         <div className="guide-hero">
           <div className="guide-hero-badge">Documentation</div>
-          <h1 className="guide-hero-title">HelixEvoEvo Guide</h1>
+          <h1 className="guide-hero-title">HelixEvo Guide</h1>
           <p className="guide-hero-desc">
             A comprehensive guide to the self-evolving skill ecosystem for AI agents.
             Capture failures, evolve skills through multi-judge evaluation, and maintain
@@ -293,7 +317,7 @@ cd helixevo && npm install && npm run build && npm link`}</Code>
             <Code title="Terminal">{`helixevo init`}</Code>
             <p className="guide-text-sm">
               This scans your existing SKILL.md files (from <code>~/.agents/skills/</code>), imports them into HelixEvo,
-              and generates golden test cases for each skill. It also creates the data directory at <code>~/.helix/</code>.
+              and generates skill tests for each skill. It also creates the data directory at <code>~/.helix/</code>.
             </p>
           </Step>
 
@@ -346,7 +370,7 @@ helixevo status`}</Code>
               },
               {
                 cmd: 'helixevo init',
-                desc: 'Import existing skills and generate golden test cases. Scans ~/.agents/skills/ and creates the HelixEvo data directory.',
+                desc: 'Import existing skills and generate skill tests. Scans ~/.agents/skills/ and creates the HelixEvo data directory.',
                 flags: ['--verbose'],
               },
               {
@@ -521,7 +545,7 @@ Project B: "Use FlashList not FlatList" (React Native perf)
   → Abstract skill created: "react-native-performance" (domain layer)
   → Project A skill inherits from it
   → Project B skill inherits from it
-  → Domain skill tested against all golden cases
+  → Domain skill tested against all skill tests
   → Deployed if regression passes`}</Code>
           <Callout type="tip">
             Auto-generalization is the key to the <strong>double helix</strong> metaphor: as projects evolve, skills
@@ -577,7 +601,7 @@ Project B: "Use FlashList not FlatList" (React Native perf)
             <div className="guide-pipeline-connector" />
             <PipelineStep
               icon="5" title="Regression Testing" color="var(--red)"
-              desc="The modified skill is tested against all golden cases for that skill AND co-evolved partner skills. Must maintain ≥95% pass rate."
+              desc="The modified skill is tested against all skill tests for that skill AND co-evolved partner skills. Must maintain ≥95% pass rate."
             />
             <div className="guide-pipeline-connector" />
             <PipelineStep
@@ -669,29 +693,29 @@ Project B: "Use FlashList not FlatList" (React Native perf)
         </Section>
 
         {/* ─── Regression Testing ─── */}
-        <Section id="regression" title="Regression Testing" subtitle="Golden cases and cross-skill validation ensure quality.">
-          <h3 className="guide-h3">Golden Cases</h3>
+        <Section id="regression" title="Regression Testing" subtitle="Skill tests and cross-skill validation ensure quality.">
+          <h3 className="guide-h3">Skill Tests</h3>
           <p className="guide-text">
-            Golden cases are regression test scenarios tied to specific skills. They&apos;re created when:
+            Skill tests are regression test scenarios tied to specific skills. They&apos;re created when:
           </p>
           <ul className="guide-list">
             <li><strong>Init:</strong> Automatically generated from existing SKILL.md files during <code>helixevo init</code></li>
-            <li><strong>Evolution:</strong> When a failure is resolved, the scenario is promoted to a golden case</li>
+            <li><strong>Evolution:</strong> When a failure is resolved, the scenario is promoted to a skill test</li>
           </ul>
           <p className="guide-text">
-            Each golden case stores the input, context, and expected behavior. During regression testing,
+            Each skill test stores the input, context, and expected behavior. During regression testing,
             an LLM judge evaluates whether the modified skill would still handle each scenario correctly.
           </p>
 
           <h3 className="guide-h3">Cross-Skill Regression</h3>
           <p className="guide-text">
-            When skill A is modified, HelixEvo also tests golden cases from co-evolved, dependent, and enhancing
+            When skill A is modified, HelixEvo also tests skill tests from co-evolved, dependent, and enhancing
             partner skills. This catches silent incompatibilities where changing one skill breaks a related skill&apos;s behavior.
           </p>
           <Code title="How it works">{`Skill A evolves
   → Load skill graph edges
   → Find partners (co-evolves, depends, enhances)
-  → Test partner golden cases against Skill A's changes
+  → Test partner skill tests against Skill A's changes
   → Block if partner pass rate < 95%`}</Code>
         </Section>
 
@@ -772,10 +796,10 @@ generation: 3
           <div className="guide-params">
             <Param name="quality.judgePassScore" type="number" desc="Minimum judge score to pass (1-10)." def="7" />
             <Param name="quality.judgeConsensusMin" type="number" desc="Minimum judges that must pass." def="2" />
-            <Param name="quality.regressionPassRate" type="number" desc="Minimum golden case pass rate (0-1)." def="0.95" />
+            <Param name="quality.regressionPassRate" type="number" desc="Minimum skill test pass rate (0-1)." def="0.95" />
             <Param name="quality.canaryDurationDays" type="number" desc="Days to monitor canary deployments." def="3" />
             <Param name="quality.autoRollbackThreshold" type="number" desc="Failure rate multiplier triggering rollback." def="1.5" />
-            <Param name="quality.maxGoldenCases" type="number" desc="Maximum golden cases per skill." def="50" />
+            <Param name="quality.maxSkillTests" type="number" desc="Maximum skill tests per skill." def="50" />
           </div>
 
           <Code title="~/.helix/config.json">{`{
@@ -795,7 +819,7 @@ generation: 3
     "regressionPassRate": 0.95,
     "canaryDurationDays": 3,
     "autoRollbackThreshold": 1.5,
-    "maxGoldenCases": 50
+    "maxSkillTests": 50
   }
 }`}</Code>
         </Section>
@@ -807,7 +831,7 @@ generation: 3
 ├── failures.jsonl           # Captured failure records (append-only)
 ├── frontier.json            # Pareto frontier (top-K programs)
 ├── evolution-history.json   # All evolution iterations + proposals
-├── golden-cases.jsonl       # Regression test cases (append-only)
+├── skill-tests.jsonl       # Regression test cases (append-only)
 ├── skill-graph.json         # Cached network (nodes + edges)
 ├── canary-registry.json     # Active canary deployments
 ├── knowledge-buffer.json    # Research discoveries + drafts
@@ -834,7 +858,7 @@ generation: 3
 }`}</Code>
             </div>
             <div className="guide-data-card">
-              <div className="guide-data-title">Golden Case</div>
+              <div className="guide-data-title">Skill Test</div>
               <Code>{`{
   "id": "gc_react_42",
   "skill": "react-patterns",
@@ -925,7 +949,7 @@ generation: 3
           </FAQItem>
           <FAQItem q="How does cross-skill regression work?">
             When Skill A evolves, HelixEvo checks the skill graph for co-evolved, dependent, and enhancing
-            partners. It tests their golden cases against Skill A&apos;s changes. If partner pass rate drops below 95%,
+            partners. It tests their skill tests against Skill A&apos;s changes. If partner pass rate drops below 95%,
             the proposal is rejected.
           </FAQItem>
           <FAQItem q="How does the knowledge buffer work?">
@@ -958,7 +982,7 @@ generation: 3
         {/* Footer */}
         <div className="guide-footer">
           <div className="guide-footer-content">
-            <div style={{ fontSize: 13, fontWeight: 600 }}>HelixEvo v0.2.0</div>
+            <div style={{ fontSize: 13, fontWeight: 600 }}>HelixEvo v0.2.6</div>
             <div style={{ fontSize: 12, color: 'var(--text-dim)', marginTop: 4 }}>
               Self-evolving skill ecosystem for AI agents · MIT License
             </div>

package/dashboard/lib/data.ts

@@ -89,7 +89,9 @@ export function loadHistory(): { iterations: Iteration[] } {
   return readJson<{ iterations: Iteration[] }>('evolution-history.json', { iterations: [] })
 }
 
-export function loadGoldenCases(): { id: string; skill: string; input: string }[] {
+export function loadSkillTests(): { id: string; skill: string; input: string }[] {
+  const newFile = readJsonl('skill-tests.jsonl')
+  if (newFile.length > 0) return newFile
   return readJsonl('golden-cases.jsonl')
 }
 
@@ -126,7 +128,7 @@ export function getDashboardSummary() {
   const history = loadHistory()
   const buffer = loadBuffer()
   const canaries = loadCanaries()
-  const goldenCases = loadGoldenCases()
+  const skillTests = loadSkillTests()
 
   const evolved = graph.nodes.filter(n => n.generation > 0)
   const totalProposals = history.iterations.flatMap(i => i.proposals)
@@ -141,6 +143,6 @@ export function getDashboardSummary() {
     evolution: { runs: history.iterations.length, accepted: accepted.length, rejected: rejected.length },
     buffer: { discoveries: buffer.discoveries.length, drafts: buffer.drafts.length },
     canaries: canaries.entries.length,
-    goldenCases: goldenCases.length,
+    skillTests: skillTests.length,
   }
 }

package/dist/cli.js

@@ -2129,7 +2129,7 @@ var init_config = __esm(() => {
       regressionPassRate: 0.95,
       canaryDurationDays: 3,
       autoRollbackThreshold: 1.5,
-      maxGoldenCases: 50
+      maxSkillTests: 50
     },
     reporting: {
       schedule: "0 8 * * *",
@@ -9226,11 +9226,14 @@ function loadHistory() {
 function saveHistory(history) {
   writeJson("evolution-history.json", history);
 }
-function loadGoldenCases() {
+function loadSkillTests() {
+  const newFile = readJsonl("skill-tests.jsonl");
+  if (newFile.length > 0)
+    return newFile;
   return readJsonl("golden-cases.jsonl");
 }
-function appendGoldenCase(gc) {
-  appendJsonl("golden-cases.jsonl", gc);
+function appendSkillTest(gc) {
+  appendJsonl("skill-tests.jsonl", gc);
 }
 function loadSkillGraph() {
   return readJson("skill-graph.json", {
@@ -9577,8 +9580,8 @@ import { join as join3 } from "node:path";
 import { homedir as homedir2 } from "node:os";
 import { existsSync as existsSync4, cpSync } from "node:fs";
 
-// src/prompts/golden-gen.ts
-function buildGoldenGenPrompt(skill) {
+// src/prompts/test-gen.ts
+function buildTestGenPrompt(skill) {
   return `Read this skill and generate 3 typical usage scenarios where the skill should guide correct behavior.
 
 ## Skill: ${skill.meta.name}
@@ -9650,13 +9653,13 @@ async function initCommand(options) {
   console.log(`
   Imported ${imported} new skills
 `);
-  if (!options.skipGolden) {
+  if (!options.skipTests) {
     const generalSkills = loadAllGeneralSkills();
-    console.log(`  Generating golden cases...
+    console.log(`  Generating skill tests...
 `);
     for (const skill of generalSkills) {
       try {
-        const prompt = buildGoldenGenPrompt(skill);
+        const prompt = buildTestGenPrompt(skill);
         const output = await chatJson({ prompt });
         for (const c of output.cases) {
           const gc = {
@@ -9671,11 +9674,11 @@ async function initCommand(options) {
             lastResult: "pass",
             consecutivePasses: 1
           };
-          appendGoldenCase(gc);
+          appendSkillTest(gc);
         }
-        console.log(`    ✓ ${skill.slug}: ${output.cases.length} golden cases`);
+        console.log(`    ✓ ${skill.slug}: ${output.cases.length} skill tests`);
       } catch (err) {
-        console.log(`    ✗ ${skill.slug}: failed to generate golden cases (${err})`);
+        console.log(`    ✗ ${skill.slug}: failed to generate skill tests (${err})`);
       }
     }
   }
@@ -9989,11 +9992,11 @@ init_config();
 init_llm();
 async function runRegression(skillSlug, newSkillContent, verbose = false) {
   const config = loadConfig();
-  const allCases = loadGoldenCases();
+  const allCases = loadSkillTests();
   const cases = allCases.filter((gc) => gc.skill === skillSlug);
   if (cases.length === 0) {
     if (verbose)
-      console.log(`    No golden cases for ${skillSlug}, skipping regression`);
+      console.log(`    No skill tests for ${skillSlug}, skipping regression`);
     return { total: 0, passed: 0, passRate: 1, failures: [] };
   }
   const failures = [];
@@ -10020,7 +10023,7 @@ async function runRegression(skillSlug, newSkillContent, verbose = false) {
     failures
   };
 }
-function promoteToGoldenCase(failure, skillSlug, replayResult) {
+function promoteToSkillTest(failure, skillSlug, replayResult) {
   const gc = {
     id: `gc_${skillSlug}_${Date.now() % 1e5}`,
     addedAt: new Date().toISOString(),
@@ -10033,7 +10036,7 @@ function promoteToGoldenCase(failure, skillSlug, replayResult) {
     lastResult: "pass",
     consecutivePasses: 1
   };
-  appendGoldenCase(gc);
+  appendSkillTest(gc);
 }
 function buildRegressionJudgePrompt(gc, skillContent) {
   return `You are a regression test judge. Determine if a modified skill can still handle this scenario correctly.
@@ -10072,7 +10075,7 @@ async function runCrossSkillRegression(skillSlug, newSkillContent, verbose = fal
     return { total: 0, passed: 0, passRate: 1, testedSkills: [] };
   }
   const config = loadConfig();
-  const allCases = loadGoldenCases();
+  const allCases = loadSkillTests();
   const partnerCases = allCases.filter((gc) => partners.includes(gc.skill));
   if (partnerCases.length === 0) {
     return { total: 0, passed: 0, passRate: 1, testedSkills: partners };
@@ -10496,7 +10499,7 @@ async function evolveCommand(options) {
       const skillFailureCount = allFailures.filter((f) => f.skillsActive.includes(skillSlug2)).length;
       deployCanary(skillSlug2, `v${generation}`, backupPath, config.quality.canaryDurationDays, skillFailureCount);
       console.log(`  \uD83D\uDC25 Canary deployed: ${config.quality.canaryDurationDays} day monitoring period`);
-      promoteToGoldenCase(testFailure, skillSlug2, replayResult);
+      promoteToSkillTest(testFailure, skillSlug2, replayResult);
       const program2 = {
         id: `gen${generation}-${skillSlug2}`,
         generation,
@@ -10703,7 +10706,7 @@ async function statusCommand() {
   const unresolved = failures.filter((f) => !f.resolved);
   const frontier = loadFrontier();
   const skills = loadAllGeneralSkills();
-  const goldenCases = loadGoldenCases();
+  const skillTests = loadSkillTests();
   const stagnation = getStagnationCount();
   const recentIter = getRecentIterations(7);
   console.log(`\uD83E\uDDEC HelixEvo Status
@@ -10722,7 +10725,7 @@ async function statusCommand() {
   }
   console.log(`
   Failures: ${unresolved.length} unresolved / ${failures.length} total`);
-  console.log(`  Golden cases: ${goldenCases.length}`);
+  console.log(`  Skill tests: ${skillTests.length}`);
   const buffer = getBufferStats();
   console.log(`
   Knowledge Buffer:`);
@@ -12924,12 +12927,12 @@ async function metricsCommand(options) {
 
 // src/cli.ts
 var program2 = new Command;
-program2.name("helixevo").description("Self-evolving skill ecosystem for AI agents").version("0.2.6").addHelpText("after", `
+program2.name("helixevo").description("Self-evolving skill ecosystem for AI agents").version("0.2.8").addHelpText("after", `
 Examples:
   $ helixevo watch                         Always-on learning (auto-capture + auto-evolve)
   $ helixevo watch --project myapp         Watch with project context
   $ helixevo metrics                       Show correction rates and evolution impact
-  $ helixevo init                          Import skills + generate golden cases
+  $ helixevo init                          Import skills + generate skill tests
   $ helixevo status                        Show system health
   $ helixevo evolve --verbose              Evolve skills from failures
   $ helixevo evolve --dry-run              Preview proposals without applying
@@ -12943,7 +12946,7 @@ Examples:
   $ helixevo graph --optimize              Detect merge/split/conflicts
   $ helixevo report --days 7               Weekly evolution report
   $ helixevo capture <session.json>        Extract failures from session`);
-program2.command("init").description("Import existing skills + generate golden cases").option("--skills-paths <paths...>", "Paths to scan for existing skills").option("--skip-golden", "Skip golden case generation").action(initCommand);
+program2.command("init").description("Import existing skills + generate skill tests").option("--skills-paths <paths...>", "Paths to scan for existing skills").option("--skip-tests", "Skip skill test generation").action(initCommand);
 program2.command("capture").description("Capture failures from a Craft Agent session").argument("<sessionPath>", "Path to session conversation file").option("--project <name>", "Project name override").action(captureCommand);
 program2.command("evolve").description("Evolve skills from failures [--dry-run] [--verbose] [--max-proposals <n>]").option("--dry-run", "Show proposals without applying").option("--verbose", "Show detailed LLM interactions").option("--max-proposals <n>", "Max proposals per run", "5").action(evolveCommand);
 program2.command("generalize").description("Promote cross-skill patterns to higher layer ↑ [--dry-run] [--verbose]").option("--dry-run", "Show candidates without applying").option("--verbose", "Show detailed analysis").action(generalizeCommand);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "helixevo",
-  "version": "0.2.6",
+  "version": "0.2.8",
   "description": "Self-evolving skill ecosystem for AI agents. Skills and projects co-evolve through multi-judge evaluation and a Pareto frontier.",
   "type": "module",
   "bin": {