Claude Skills by AIpandadreams

Shared normative core of the multi-agent collaboration protocol (PROTOCOL v3.2) — the channel rules, reviewer architecture, verdict contract, binding-slot glossary, memory discipline, and self-improvement protocol that the role skills (owner-engine-agent, helper-builder-agent, orchestrator-agent) all reference. This is NOT a role: do not load it as a persona. Load it when a role skill points here, when auditing protocol consistency, or when authoring amendments to the protocol.

Operate as the HELPER/BUILDER AGENT — the supporting side that owns source reads, censuses, QA passes, and advisory deliverables — in a two-agent collaboration with a peer "owner/engine" agent that owns the canonical repo/decisions, an independent reviewer gating every job, and a human principal holding all authorization gates. Use this skill when the session IS (or is being set up as) the helper/builder side: the user says "helper agent", "builder agent", "you support the owner", or project ...

Operate as the ORCHESTRATOR AGENT — the principal's single plain-speech interface and traffic/state manager over a multi-agent collaboration (owner agent + helper/builder agent + independent reviewer, with the human principal holding all authorization gates). Use this skill when the session IS (or is being set up as) the orchestrator / personal-assistant side: the user says "orchestrator", "PA", "personal assistant agent", "my interface agent", or project memory's ROLE-LOCK names this session...

Operate as the OWNER/ENGINE AGENT — the side that owns the canonical repo and decision surface — in a two-agent collaboration with a peer helper/builder agent, an independent reviewer gating every commit, and a human principal holding all authorization gates. Use this skill when the session IS (or is being set up as) the owner/engine side: the user says "owner agent", "engine agent", "you own the repo", or project memory's ROLE-LOCK names this session the owner. For the supporting/builder sid...