name	discover
description	Trigger AI Discovery chain for exploring and optimizing prompts, skills, and agents through multi-agent debate
trigger	When user says "discover", "explore possibilities", "optimize prompts", "improve agents", "brainstorm with agents"

AI Discovery Chain

Launch multi-agent exploration and debate to discover novel solutions and optimize AI artifacts.

Usage

/discover <topic>
/discover optimize the marketing template for viral content
/discover improve the explorer agent's creativity
/discover find better patterns for error handling

What It Does

Phase 0: Intent Discovery — Clarifies what you're trying to achieve
Phase 1: Parallel Exploration — Multiple agents explore from different angles
Phase 2: Forum Debate — Agents challenge and stress-test ideas
Phase 3: Synthesis — Moderator extracts actionable insights
Phase 4: Optimization — Apply improvements to prompts/skills/agents
Phase 5: Validation — Verify changes work correctly
Phase 6: Report — Document discoveries and changes

Chain Flow

intent-discovery → parallel-exploration → forum-debate → synthesis
                           ↑                               ↓
                           └──── (max 2 rounds) ←──────────┘
                                                           ↓
                                              optimization → validation → report

Execution

When user invokes /discover:

Parse the topic from arguments
Spawn intent-discoverer agent with the topic to clarify intent and define exploration directive
Wait for intent analysis — this shapes all subsequent exploration
The chain continues automatically via hooks

Spawn Command

Agent(
  name: "intent-analyst",
  subagent_type: "intent-discoverer",
  prompt: "Analyze intent for: {user_topic}. Define exploration directive, success criteria, and what to avoid."
)

Options

/discover <topic> — Full chain with intent discovery
/discover fast <topic> — Skip intent discovery, jump to exploration
/discover debate <ideas> — Skip exploration, go straight to debate on provided ideas

Output

The chain produces:

Discovery Report — Summary of exploration, debates, and conclusions
Optimized Artifacts — Improved agent/skill/prompt files (if optimization phase runs)
Seeds for Future — Unresolved questions for next exploration cycle

Examples

/discover how to make the critic agent more constructive
/discover ways to reduce token usage in the development chain
/discover fast viral marketing hooks for tech products
/discover debate "use websockets vs SSE for real-time updates"

Más de este repositorio

mismo repositorio

brainbrew-codex-setup

brainbrewlabs/brainbrew-devkit

Use when setting up, verifying, or repairing BrainBrew DevKit in Codex, including hooks, global skills, MCP config, runner files, and plugin-native assets.

2026-05-177

brainbrew-mcp-guidance

brainbrewlabs/brainbrew-devkit

Use when configuring or troubleshooting BrainBrew MCP access in Codex, including .mcp.json, codex mcp commands, and available BrainBrew MCP tools.

2026-05-177

brainbrew-workflow-recipes

brainbrewlabs/brainbrew-devkit

Use when the user wants to run or follow a BrainBrew workflow in Codex, including develop, review, docs, devops, research, support, data, moderation, marketing, or skill-dev recipes.

2026-05-177

chain-builder

brainbrewlabs/brainbrew-devkit

Set up workflow templates and manage chain flows. Trigger when user says: "set up a workflow", "create a development workflow", "I need a CI/CD pipeline", "set up devops", "bump develop", "bump devops", "add agent to chain", "show chain flow", "edit chain config", "create custom workflow", "add a team node", "run agents in parallel", "create agent team", "validate chain", "check chain config", "verify chain".

2026-05-157

docs

brainbrewlabs/brainbrew-devkit

Build and manage VitePress documentation site. Trigger when user says: "build docs", "deploy docs", "preview docs", "update docs", "docs dev server", "run docs locally", "check docs build".

2026-03-297

skill-finder

brainbrewlabs/brainbrew-devkit

Search and install skills from multiple sources (Vercel Skills, GitHub, Anthropic official). Use when user says "find skill", "search skill", "install skill", or when agent-improver needs to find skills for an agent.

2026-03-297

name	discover
description	Trigger AI Discovery chain for exploring and optimizing prompts, skills, and agents through multi-agent debate
trigger	When user says "discover", "explore possibilities", "optimize prompts", "improve agents", "brainstorm with agents"

AI Discovery Chain

Launch multi-agent exploration and debate to discover novel solutions and optimize AI artifacts.

Usage

/discover <topic>
/discover optimize the marketing template for viral content
/discover improve the explorer agent's creativity
/discover find better patterns for error handling

What It Does

Phase 0: Intent Discovery — Clarifies what you're trying to achieve
Phase 1: Parallel Exploration — Multiple agents explore from different angles
Phase 2: Forum Debate — Agents challenge and stress-test ideas
Phase 3: Synthesis — Moderator extracts actionable insights
Phase 4: Optimization — Apply improvements to prompts/skills/agents
Phase 5: Validation — Verify changes work correctly
Phase 6: Report — Document discoveries and changes

Chain Flow

intent-discovery → parallel-exploration → forum-debate → synthesis
                           ↑                               ↓
                           └──── (max 2 rounds) ←──────────┘
                                                           ↓
                                              optimization → validation → report

Execution

When user invokes /discover:

Parse the topic from arguments
Spawn intent-discoverer agent with the topic to clarify intent and define exploration directive
Wait for intent analysis — this shapes all subsequent exploration
The chain continues automatically via hooks

Spawn Command

Agent(
  name: "intent-analyst",
  subagent_type: "intent-discoverer",
  prompt: "Analyze intent for: {user_topic}. Define exploration directive, success criteria, and what to avoid."
)

Options

/discover <topic> — Full chain with intent discovery
/discover fast <topic> — Skip intent discovery, jump to exploration
/discover debate <ideas> — Skip exploration, go straight to debate on provided ideas

Output

The chain produces:

Discovery Report — Summary of exploration, debates, and conclusions
Optimized Artifacts — Improved agent/skill/prompt files (if optimization phase runs)
Seeds for Future — Unresolved questions for next exploration cycle

Examples

/discover how to make the critic agent more constructive
/discover ways to reduce token usage in the development chain
/discover fast viral marketing hooks for tech products
/discover debate "use websockets vs SSE for real-time updates"