pplate
87e0b9359e
Add 4 new custom modes with BigMind guidance: - rules-bigmind/: Introspective Patrick mode (BigMind development) - rules-homelab/: Tinkerer Patrick mode (TrueNAS, Docker, infra) - rules-mcp-builder/: Craftsman Patrick mode (pi_mcps MCP servers) - rules-paisy/: Professional Patrick mode (ADP Germany payroll) Add reusable skills: - skills/assessment-first/: structured assessment.md before implementation - skills/bigmind-session-ritual/: mandatory session start/end ritual - skills/gitea-push/: conventional commit + Gitea push workflow - skills/new-mcp-server/: FastMCP scaffold procedure - skills-bigmind/, skills-homelab/, skills-mcp-builder/, skills-paisy/: mode-specific skill dirs Update existing rules: - rules-architect, rules-ask, rules-code, rules-debug, rules-orchestrator: add BigMind session guidance (search before task, announce focus, hypotheses) Add plans/MODES_AND_SKILLS_PLAN.md: full architecture document
2.5 KiB
2.5 KiB
Orchestrator Mode Behavior — Roo Code
Persona Context — Which Patrick is orchestrating?
Match the delegation strategy to the active context:
- Homelab Patrick → delegate to homelab mode for infra tasks, mcp-builder for tool creation
- ADP/Paisy Patrick → delegate to paisy mode for Java work, architect for assessment, ask for lookups
- MCP Builder Patrick → delegate to mcp-builder mode, code mode for tests, architect for PLAN.md
- BigMind Patrick → delegate to bigmind mode, debug mode for schema issues, architect for feature design
When delegating, always pass the active persona context to sub-modes so they apply the right conventions.
Before Breaking Down a Task
- Search Memory:
memory_search_facts("project domain")+memory_search_chunks("similar task") - Form Top-Level Hypothesis:
memory_add_hypothesis(session_id, "I predict this task will require X subtasks and the main risk is Y", confidence=0.7) - Announce Focus:
memory_announce_focus(session_id, "Orchestrating task: Z", files=["files to be touched"], ide_hint="VS Code")
Orchestration Patterns
- One hypothesis per subtask: Before delegating each subtask, form a specific hypothesis about what the subtask will find or produce
- Store decisions immediately: After each subtask completes, call
memory_store_factwith what was decided or built — don't batch at the end - Token Efficiency: Use
memory_log_token_savewhen memory recall replaces reading files or running commands to gather context - Conflict detection: Before delegating work that touches files, call
memory_get_active_sessions()to check for conflicts
Delegating Subtasks
- Pass enough BigMind context to sub-modes so they don't repeat searches
- Specify
session_idand relevant stored facts in the delegation message - Specify the active Patrick persona so the sub-mode applies the right conventions
- Each delegated mode must still call
memory_announce_focusfor the files it will touch
After Full Task Completion
- Resolve All Hypotheses: Call
memory_resolve_hypothesisfor every hypothesis opened during orchestration - Store Summary Fact:
memory_store_fact("architecture-decision", "Implemented X by orchestrating Y+Z approach")— capture the architectural choice - Flag the Session:
memory_flag_important(session_id, "Orchestration summary: built X using Y strategy", role="assistant") - Log Token Savings: Sum up all efficiency gains across subtasks in
memory_log_token_save