ai-council_skill
- C#
0
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill mkalhitti-cloud/universal-or-strategy --skill ai-council- debate-template.md1.0 KB
- SKILL.md9.9 KB
Overview
This skill enables structured multi-AI collaboration by having multiple agents write directly into a shared debate file to debate, rebut, and reach a consensus on high-impact decisions. It removes manual copy/paste, enforces three-round debate structure, and is optimized for major trading or architectural choices that require diverse AI perspectives. Use it to get clear, documented recommendations and an auditable decision trail.
How this skill works
You create a debate file in a designated folder and seed it with context, the decision question, and a round-by-round template. Each AI reads the full file, finds its named section, and appends its position, rebuttal, and final stance across three rounds. As moderator, you call rounds, synthesize the final recommendation, and record action items and risks.
When to use it
- Major strategy decisions (e.g., add a new trading strategy)
- High-stakes parameter changes that could materially affect P&L
- Significant architecture or scaling decisions (multiple accounts, new infrastructure)
- Emergency protocol choices (disconnect handling, failover plans)
- When you need diverse AI viewpoints to avoid single-agent blind spots
Best practices
- Provide clear, concise context and constraints at the top of the debate file
- Limit each AI response (word or bullet limits) to keep rounds focused
- Enforce three rounds: initial positions, rebuttals, and consensus-building
- Require AIs to reference other agents' points and ask targeted questions
- Archive completed debates for learning and post-mortem analysis
- If progress stalls, add a focused extra round asking "what would change your mind?"
Example use cases
- Deciding whether to implement RMA in the current release vs. postpone to a major version
- Setting risk-management thresholds that could change potential losses over $1,000
- Choosing failover and reconnection behavior for trading data feeds
- Agreeing on rollout strategy: feature flag + sim testing vs full deployment
- Resolving conflicting architecture proposals when building multi-strategy support
FAQ
A full three-round debate takes about 30–60 minutes if each AI responds promptly and you, as moderator, synthesize at the end.
What if AIs don't converge by Round 3?
Add a Round 4 focused on change drivers, pose a decisive tie-breaker question, or propose parallel tests and choose by empirical results.