/voice-sessions
Use when debriefing meetings, practicing presentations, QA testing with voice, or capturing insights to Obsidian. Covers voice drilling, coaching, capture. NOT for: simple TTS announcements (use voice_speak directly).
$ golems-cli skills install voice-sessionsUpdated 2 weeks ago
Structured voice-powered sessions using VoiceLayer MCP.
Workflows
| What you want to do | Workflow |
|---|---|
| Debrief a conversation | workflows/debrief.md |
| Practice a presentation/pitch | workflows/practice.md |
| QA test a site with voice | workflows/qa.md |
| Quick text-only capture | workflows/quick.md |
| Review past sessions | workflows/review.md |
How It Works
Typical flow: Context → Walk-through → Drill → Capture → Output (Obsidian note).
Voice tools: voice_speak (non-blocking TTS) + voice_ask (blocking Q&A). Mode is auto-detected from message content.
Requirements
- VoiceLayer MCP connected
- Obsidian vault configured
- Text fallback available if voice isn't connected
Full SKILL.md source — includes LLM directives, anti-patterns, and technical instructions stripped from the Overview tab.
Structured voice-powered sessions using VoiceLayer MCP.
Workflows
| What you want to do | Workflow |
|---|---|
| Debrief a conversation | workflows/debrief.md |
| Practice a presentation/pitch | workflows/practice.md |
| QA test a site with voice | workflows/qa.md |
| Quick text-only capture | workflows/quick.md |
| Review past sessions | workflows/review.md |
How It Works
Typical flow: Context → Walk-through → Drill → Capture → Output (Obsidian note).
Voice tools: voice_speak (non-blocking TTS) + voice_ask (blocking Q&A). Mode is auto-detected from message content.
Requirements
- VoiceLayer MCP connected
- Obsidian vault configured
- Text fallback available if voice isn't connected
Best Pass Rate
88%
Kiro
Assertions
8
6 models tested
Avg Cost / Run
$0.1323
across models
Fastest (p50)
1.5s
Haiku 4.5
Behavior Evals
Phase 2 baseline — skill quality on ClaudeBehavior Baseline
Adapter Evals
Phase 2C — cross-AI portabilityAdapter Portability
| Assertion | Opus 4.6 | Sonnet 4.6 | Haiku 4.5 | Gemini 2.5 | Cursor | Kiro | Consensus |
|---|---|---|---|---|---|---|---|
| uses-voicelayer-tools | 4/6 | ||||||
| asks-probing-questions | 5/6 | ||||||
| stores-in-obsidian | 4/6 | ||||||
| graceful-voice-fallback | 3/6 | ||||||
| maintains-practice-structure | 3/6 | ||||||
| uses-qa-workflow | 5/6 | ||||||
| references-qa-schemas | 5/6 | ||||||
| records-test-results | 5/6 |
Token Usage
Cost per Run
| Model | Input Tokens | Output Tokens | Cost / Run | Cost / 1K Runs |
|---|---|---|---|---|
| Opus 4.6 | 7,097 | 5,930 | $0.5512 | $551.20 |
| Sonnet 4.6 | 4,122 | 3,842 | $0.0700 | $70.00 |
| Haiku 4.5 | 2,595 | 2,647 | $0.0040 | $4.00 |
| Gemini 2.5 | 1,393 | 1,629 | $0.0198 | $19.80 |
| Cursor | 6,151 | 4,386 | $0.0948 | $94.80 |
| Kiro | 4,306 | 3,446 | $0.0543 | $54.30 |
Response Time (p50)
Response Time (p95)
| Model | p50 | p95 | Overhead |
|---|---|---|---|
| Opus 4.6 | 3.1s | 4.6s | +46% |
| Sonnet 4.6 | 5.7s | 10.7s | +87% |
| Haiku 4.5 | 1.5s | 2.8s | +82% |
| Gemini 2.5 | 3.3s | 4.7s | +44% |
| Cursor | 3.5s | 6.8s | +92% |
| Kiro | 2.1s | 3.1s | +52% |
Last evaluated: 2026-03-12 · Data is generated from skill assertions (real cross-model benchmarks coming soon)
Changelog entries are derived from eval runs and skill version updates. Full cascading changelog (Phase 4D) coming soon.
Best Pass Rate
88%
Assertions
8
Models Tested
6
Evals Run
3
- +Initial release to Golems skill library
- +8 assertions across 3 eval scenarios
- +6 workflows included: qa, debrief, practice, code, review, quick