bicameral-guided

Block agent actions when bicameral detects drifted or divergent decisions.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/BicameralAI/bicameral-mcp --skill bicameral-guided
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bicameral-guided
Source: https://github.com/BicameralAI/bicameral-mcp/tree/main/.claude/skills/bicameral-guided
Command: npx skills add https://github.com/BicameralAI/bicameral-mcp --skill bicameral-guided

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Guided mode adds a strict-enforcement layer on top of bicameral.search and bicameral.brief, stopping the agent when drifted decisions, divergent decision pairs, or unresolved open questions are detected to preserve alignment with recorded intent.

Core Features & Use Cases

  • Opt-in Guided mode: blocking behavior with imperative tone for critical-path work; normal mode remains advisory.
  • Onboarding, demos, and evaluation: surfaces discrepancies before any write operation to prevent regressions.
  • Config parity: durable setup via .bicameral/config.yaml and runtime override via BICAMERAL_GUIDED_MODE.

Quick Start

Enable guided mode to make Bicameral stop the agent on drift, unresolved questions, or divergent decisions before any write operation.

Frequently Asked Questions about bicameral-guided

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I block AI agent actions when code drifts from recorded intent?

To block agent actions on decision drift, enable guided mode to enforce strict alignment between recorded intent and code, stopping operations before writes when divergent decisions or unresolved open questions are detected.

How does guided mode stop decision drift during critical-path development?

Guided mode stops decision drift by applying a strict enforcement layer over search and brief operations, blocking write actions and surfacing imperative ActionHints when divergent decision pairs or unresolved open questions are detected.

Can I configure stop-and-review behavior using environment variables?

Yes, you can configure stop-and-review behavior via runtime environment variable overrides like BICAMERAL_GUIDED_MODE, or establish durable configuration parity using the .bicameral/config.yaml file to toggle between blocking and advisory modes.

What is the best way to prevent regressions during AI agent onboarding and demos?

The best way to prevent regressions during onboarding and demos is applying stop-and-review guided mode, which surfaces discrepancies and blocks write operations before execution to preserve alignment with recorded intent.

Does guided mode support advisory behavior instead of blocking agent execution?

Yes, guided mode supports advisory behavior instead of blocking. Normal mode remains advisory, while the opt-in guided mode provides blocking behavior with an imperative tone specifically for critical-path work.

Why should I use stop-and-review mode for unresolved open questions?

You should use stop-and-review mode for unresolved open questions to preserve alignment with recorded intent, ensuring the agent surfaces discrepancies and halts before executing write operations that could introduce regressions.