agentsclimarketplace

Multi agent deliberation

Skill AVIDS2/avid-skill/skills/multi-agent-deliberation

Use for high-stakes decisions: architecture forks, build-vs-buy, competitive response, major refactors, or when the user wants multi-agent debate. Spawns diverse perspectives (or Codex second opinion) and synthesizes a recommendation. Do NOT use for trivial choices or every small implementation detail.From its SKILL.md

Install
npx -y skills add AVIDS2/avid-skill --skill multi-agent-deliberation

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.4 KB, as published. Nobody here has run it

Multi-Agent Deliberation

When a wrong call is expensive, prefer competing lenses — then decide. Optional: skip for clear choices or when the user wants speed.

When to Use

  • Build vs buy / build vs wrap competitor
  • LangGraph vs Pi vs custom orchestration (after agent-framework-choice still unclear)
  • "Should we pivot given competitor X?"
  • Large rewrite vs incremental
  • User explicitly asks for multi-agent or second opinion

Procedure

1. Frame the decision

Write:

  • Question (one sentence)
  • Constraints (time, skills, stack, distribution)
  • Success metric
  • Deadline / reversibility (one-way door?)

2. Run 2–3 lenses (parallel when possible)

Pick lenses that disagree productively:

LensFocus
User/valueDoes this help the buyer this month?
Risk/securityWhat fails in prod?
Tech debt / YAGNISimplest path (ponytail spirit)
CompetitiveHow does this position vs named alternatives?
Codex/rescueIndependent second implementation or review

Implementation options:

  • Claude Code Agent tool with distinct system prompts per lens
  • Codex plugin /rescue or /adversarial-review when available
  • Sequential self-roles only if subagents unavailable (label each role clearly)

3. Synthesize (you, main agent)

Produce:

## Decision question
## Options
## Lens votes (table)
## Recommendation
## Dissent / residual risks
## What would change the decision
## Proposed next experiment (smallest test)

Do not average into mush. Pick a recommendation. Surface dissent.

4. User owns the call

Present synthesis; user confirms. Record in docs/decisions/ or dev-log when multi-session.

Anti-Patterns

  • Five agents agreeing with the same prompt (fake diversity)
  • Deliberating UI padding color
  • Hiding the recommendation behind "it depends" with no default
  • Skipping research gates by debating vibes only — attach evidence

Related

  • product-bootstrap, research-before-build, agent-framework-choice

Keep looking

Skills are one crate of 608,666. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.