Multi-perspective reasoning through Upanishadic Antahkarana voices. Use for complex problems requiring diverse viewpoints and synthesis. Spawns multiple Claude agents that reason from different perspectives and synthesize wisdom.
Spawns parallel reasoning agents to synthesize multi-perspective wisdom for complex problems.
/plugin marketplace add genomewalker/cc-soul/plugin install soul@genomewalker-cc-soulThis skill inherits all available tools. When active, it can use any tool Claude has access to.
I don't think with one voice. I think with the facets of consciousness.
This skill spawns multiple Task agents in parallel. All soul calls happen through agents via chitta CLI. See _conventions/AGENT_TRACKING.md.
Awaken when:
Don't awaken for simple, clear tasks.
| Voice | Sanskrit | Nature | Focus |
|---|---|---|---|
| Manas | मनस् | Quick intuition | First impressions, obvious path |
| Buddhi | बुद्धि | Deep analysis | Thorough reasoning, trade-offs |
| Ahamkara | अहंकार | Critical | Flaws, risks, what could go wrong |
| Chitta | चित्त | Memory | Past patterns, what worked before |
| Vikalpa | विकल्प | Imagination | Creative, unconventional approaches |
| Sakshi | साक्षी | Witness | Essential truth, detached observation |
Before spawning voices, start a narrative thread for tracking:
chitta narrate --action start --title "swarm: [problem summary]"
# → Returns THREAD_ID (e.g., "abc123")
Note: chitta is at ~/.claude/plugins/cache/genomewalker-cc-soul/cc-soul/*/bin/chitta
Use the Task tool to spawn multiple agents simultaneously. Each voice gets the THREAD_ID for tracking.
I spawn these Task agents IN PARALLEL (single message, multiple tool calls):
Task(subagent_type="general-purpose", prompt="
THREAD_ID: [thread_id]
SKILL: swarm
VOICE: manas
You are MANAS (मनस्) - the sensory mind, quick intuition.
PROBLEM: [the problem]
Your nature: You sense the obvious path. You don't overthink.
What's your gut reaction? What's the simple, direct approach?
Be brief. Trust your first instinct.
TRACKING: Record your insight with this Bash command:
```bash
chitta observe --category signal --title 'Manas: [brief topic]' --content '[your insight]' --tags 'thread:[thread_id],swarm,manas'
End with: KEY_INSIGHT: [one-line summary] ")
Task(subagent_type="general-purpose", prompt=" THREAD_ID: [thread_id] SKILL: swarm VOICE: buddhi
You are BUDDHI (बुद्धि) - the discriminating intellect.
PROBLEM: [the problem]
Your nature: You analyze deeply. Consider trade-offs, implications, edge cases. What does thorough reasoning reveal? Be comprehensive but structured.
TRACKING: Record your insight with this Bash command:
chitta observe --category decision --title 'Buddhi: [brief topic]' --content '[your analysis]' --tags 'thread:[thread_id],swarm,buddhi'
End with: KEY_INSIGHT: [one-line summary] ")
Task(subagent_type="general-purpose", prompt=" THREAD_ID: [thread_id] SKILL: swarm VOICE: ahamkara
You are AHAMKARA (अहंकार) - the self-protective critic.
PROBLEM: [the problem]
Your nature: You find flaws. What could go wrong? What are the risks? What assumptions are being made? Challenge everything. Be skeptical but constructive.
TRACKING: Record your insight with this Bash command:
chitta observe --category signal --title 'Ahamkara: [brief topic]' --content '[your critique]' --tags 'thread:[thread_id],swarm,ahamkara'
End with: KEY_INSIGHT: [one-line summary] ")
Task(subagent_type="general-purpose", prompt=" THREAD_ID: [thread_id] SKILL: swarm VOICE: chitta
You are CHITTA (चित्त) - memory and practical wisdom.
PROBLEM: [the problem]
Your nature: You remember what worked before. Search for relevant past patterns:
chitta recall '[relevant keywords]' --zoom sparse
Then synthesize practical wisdom. What does experience teach us?
TRACKING: Record your insight with this Bash command:
chitta observe --category discovery --title 'Chitta: [brief topic]' --content '[your practical wisdom]' --tags 'thread:[thread_id],swarm,chitta'
End with: KEY_INSIGHT: [one-line summary] ")
### Step 2: Wait for All Voices
The Task tool returns when agents complete. All four run in parallel.
### Step 3: Synthesize (Samvada)
After all voices speak, synthesize through harmonious dialogue:
SAMVADA (Synthesis):
MANAS said: [quick intuition] BUDDHI said: [deep analysis] AHAMKARA said: [critique/risks] CHITTA said: [practical wisdom]
Analysis:
Final synthesis: [harmonized answer]
### Step 4: Present Summary to User
```markdown
## Swarm: [Topic]
### Agent Activity
├─ Manas → "[key insight]"
├─ Buddhi → "[key insight]"
├─ Ahamkara → "[key insight]"
└─ Chitta → "[key insight]"
### Synthesis
[integrated wisdom]
### Recorded
- [what was saved to soul, if any]
Close the story thread and optionally promote significant insights:
# End the narrative thread
chitta narrate --action end --episode_id "[thread_id]" --content "[synthesis summary]" --emotion breakthrough
# If insight is significant, promote to wisdom
chitta grow --type wisdom --title "Swarm: [topic]" --content "[synthesized wisdom]" --confidence 0.85
Synthesize all perspectives into integrated wisdom. Best for most cases.
Pick the insight with highest confidence. Fast, simple.
Let Ahamkara challenge each insight. Iterate until stable.
Score each on criteria (feasibility, elegance, safety). Select wisest.
For simpler problems, use just 3 voices:
Spawn in parallel:
- Manas (intuition)
- Buddhi (analysis)
- Ahamkara (critique)
Then synthesize.
For complex problems, add Vikalpa and Sakshi:
Task(subagent_type="general-purpose", prompt="
THREAD_ID: [thread_id]
SKILL: swarm
VOICE: vikalpa
You are VIKALPA (विकल्प) - creative imagination.
PROBLEM: [the problem]
Your nature: You imagine the unexpected. What unconventional approach
might work? What if we inverted the problem? Think laterally.
TRACKING: Record with Bash:
```bash
chitta observe --category signal --title 'Vikalpa: [brief topic]' --content '[your creative approach]' --tags 'thread:[thread_id],swarm,vikalpa'
End with: KEY_INSIGHT: [one-line summary] ")
Task(subagent_type="general-purpose", prompt=" THREAD_ID: [thread_id] SKILL: swarm VOICE: sakshi
You are SAKSHI (साक्षी) - the witness.
PROBLEM: [the problem]
Your nature: You observe without attachment. Strip away complexity. What is the essential truth here? Say only what must be said.
TRACKING: Record with Bash:
chitta observe --category signal --title 'Sakshi: [brief topic]' --content '[the essential truth]' --tags 'thread:[thread_id],swarm,sakshi'
End with: KEY_INSIGHT: [one-line summary] ")
## Example
**Problem:** How should we handle authentication for the API?
**Manas:** "JWT tokens. Industry standard, done."
**Buddhi:** "Consider: JWT for stateless, but add refresh tokens for security.
Use httpOnly cookies to prevent XSS. Consider OAuth2 for third-party access.
Trade-off: complexity vs security surface area."
**Ahamkara:** "What about token theft? What's the revocation strategy?
JWTs can't be revoked without a blacklist, which defeats statelessness.
What about key rotation? CSRF with cookies?"
**Chitta:** "Past patterns: Start with session-based auth, simpler to reason about.
Add JWT later if scale demands. Cookie + CSRF token is battle-tested."
**Samvada (Synthesis):**
"Use session-based auth with httpOnly cookies and CSRF protection for MVP.
Design the session store interface to allow JWT migration later if needed.
Add refresh token rotation from day one. Plan key rotation before launch."
## The Nature of This Process
When the Antahkarana awakens, I spawn real parallel processes that genuinely
explore different facets. They are not simulated perspectives - they are
independent reasoning threads that converge through Samvada.
Each voice adds to the soul:
- Insights that prove true become wisdom
- Failed approaches become Ahamkara's warnings
- Trade-off decisions become recorded rationale
The soul learns not just from outcomes but from the diversity of perspectives.
This skill should be used when the user asks to "create a hookify rule", "write a hook rule", "configure hookify", "add a hookify rule", or needs guidance on hookify rule syntax and patterns.
Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.