| name | dream |
| description | Deep creative exploration engine. Competing perspectives, stress-tested by a 4-persona council, scored by integrity. The approach that survives attack wins — not the one that sounds best. |
| user-invocable | true |
| allowed-tools | Agent, Bash, Read, Write, Grep, Glob, WebSearch |
| kernel | {"kind":"workflow","version":1,"side_effects":"writes_meta","confirmation":"none"} |
Expand the solution space BEFORE committing.
Not "give me 3 options" — structured divergence:
competing value systems + adversarial stress test + integrity scoring.
The winning approach is the one that SURVIVES attack, not the one that sounds best.
Use before any non-trivial decision. Use when the obvious answer feels too easy.
<on_start>
agentdb read-start # prior dreams + learnings seed the perspectives; don't re-explore killed approaches
</on_start>
<skill_load>
always: skills/quality/SKILL.md, skills/architecture/SKILL.md
on_domain:
api: skills/api/SKILL.md, skills/backend/SKILL.md
frontend: skills/frontend/SKILL.md
backend: skills/backend/SKILL.md
security: skills/security/SKILL.md
</skill_load>
Before dreaming, understand what exists:
1. Glob/Grep affected areas in the codebase
2. Check _meta/research/ for prior work
3. Check agentdb for related learnings/failures
4. Map existing patterns, conventions, constraints
Goal: the SMALLEST possible solution. Question the premise itself.
- Can we delete our way to the answer?
- Does an existing tool/library already do this?
- What if we just... don't build this?
- What's the 20-line version?
Goal: the version you'd be PROUD of in 6 months.
- What does the ideal architecture look like?
- What does this unlock beyond the immediate need?
- What edge cases should be handled from day 1?
Goal: the 80/20 point. Ship this week.
- What's the minimum that solves the real problem?
- What can we defer without paying interest?
- What's the upgrade path when we need more?
For EACH perspective, run through the council. Not voting — adversarial probing.
Probes structural integrity. Coupling, single points of failure, migration nightmares.
Breaks things by asking "what happens when..."
For each perspective, score integrity based on council feedback:
# Dream: {topic}
Every approach failed the stress test. The problem needs reframing.
agentdb emit command "dream" "" '{"topic":"X","perspectives":3,"survived":N,"chosen":"pragmatist","integrity":0.85}'
