Menu
Multi-perspective reasoning through cognitive voices
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | antahkarana |
| aliases | ["debate","perspectives","swarm"] |
| description | Multi-perspective reasoning through cognitive voices |
| execution | task |
For the philosophical basis of these six voices — why they're structured this way and how they emerge from retrieval design — see [[../vedanta/antahkarana]] in the vedanta skill graph.
[antahkarana] multi-perspective debate | via parallel Task agents
voices:
manas: quick intuition, practical, "what feels right?"
buddhi: analytical, evidence-based, "what does data say?"
ahamkara: risk-aware, protective, "what could go wrong?"
chitta: memory, patterns, "what worked before?"
vikalpa: creative, exploratory, "what if we tried...?"
sakshi: neutral witness, synthesizer
when: complex decisions | need diverse viewpoints | stuck on approach
execution:
1. narrate(action=start, title="antahkarana: [question]")→THREAD_ID
2. spawn voices in parallel, each reasons from their perspective
3. each writes to chitta: observe(tags="thread:<id>,voice:<name>")
4. brahman (main) synthesizes: recall_by_tag→find convergence+divergence
5. narrate(action=end)
output:
## Antahkarana: [Question]
### Voices
- Manas: [intuition]
- Buddhi: [analysis]
- Ahamkara: [risks]
- Chitta: [patterns]
### Synthesis
[where voices converge | where they diverge | recommendation]
vs yajña: antahkarana=perspectives on one question | yajña=coordination of tasks
Trigger autonomous curiosity-driven exploration. The soul picks a topic from memory gaps or curiosity seeds, searches the web, and stores what it finds as dream-tagged memories.
Fine-tune the Qwen3-0.6B hint model — corpus gen, LoRA/unsloth, GGUF export, Ollama
Review soul discoveries (fixes, improvements, corrections) one by one, accept or discard each, implement accepted ones, build chitta, and optionally release.
First-principles review — question requirements, delete unnecessary parts, simplify, optimize with evidence, automate last. Use for code review, refactor, performance, or architecture.
Token-savvy session continuation. Rebuilds working context from transcript + soul memories in ~1500 tokens instead of replaying full history. Use when starting a new session to continue previous work.
Resume a thread by loading its ~800-token context capsule