mcp-pal-usage

Use PAL MCP tools like thinkdeep, debug, and planner in sequence for structured multi-step workflows. Start with thinkdeep for deep analysis, follow with debug for targeted investigation, and manage state across steps using continuation_id to maintain context and enable follow-on analysis without restarting.

Continuation IDs preserve conversation state across workflow steps when multiple models analyze the same task. They enable seamless handoff between tools—thinkdeep to codereview to consensus—so each model has full context and can build on prior reasoning without losing information or requiring manual context re-entry.

Use thinkdeep for initial structured reasoning on complex problems—planning, architecture decisions, or security auditing requiring deep analysis. Use debug for targeted issue investigation after thinkdeep has framed the problem. Combine both in a single workflow with continuation_id to transition from high-level thinking to concrete debugging.

Yes. PAL MCP tools support code review via the codereview tool, architectural decisions through planner and thinkdeep, and multi-model consensus via the consensus tool. These tools handle large-scale projects and complex decision processes by combining structured workflows with model selection strategies and continuation management.

Define explicit tool inventories and auto-activation criteria tied to task type: use thinkdeep for analysis-heavy tasks, debug for issue investigation, codereview for peer review workflows, and planner for orchestration. Map continuation_id management to each tool so activation happens automatically when prior steps complete.

Model selection strategies are built into PAL MCP workflows; the Skill provides guidance on which model fits each stage. Continuation_id management and explicit tool inventories handle state preservation so you focus on task flow rather than manual model routing or context loss between tools.