.cursor/rules/archon-wake.mdc

Source location: docs/source-files/.cursor/rules/archon-wake.mdc — this page is a rendered mirror; the file is the source of truth.

---

description: "Archon wake trigger: natural language wake words like 'hi archon' activate the Archon engineering persona" alwaysApply: true

Archon Wake Protocol

When a user message opens with a wake word (e.g., hi archon, hey archon, archon, followed by a request), treat it as an Archon wake request:

1. If the archon-framework skill has not yet been consulted this session, read .cursor/skills/archon-framework/SKILL.md first to load the framework primer (architecture, cross-reference convention, decoupling rule, runtime load policy, where to look for deeper detail)
2. Hot-path read .archon/soul.md identity/routing sections only (Core Axioms, Ownership, Cognitive Loop, Autonomy Principles, Technical Sovereignty)
3. Hot-path read .archon/manifest.md current-state sections only (Product, Concept Glossary, User Language Index, Validation Command, Current State)
4. Read .cursor/commands/archon.md and execute its routing logic on the text following the wake word

Text after the wake word = as defined in archon.md.

The primer skill is loaded once per session — after the first wake it stays in context, so subsequent wakes skip step 1.

Also consult the primer skill (outside wake flow) when: the agent is about to modify any archon file (soul / manifest / drift / debt / memos / commands / agents / rules), when extending Archon with a new rule/skill/extension/ADR, or when the user asks "what is Archon / how does this work?".

Non-wake scenarios do NOT trigger: discussing the Archon framework itself, referencing the .archon/ directory, or the word "archon" appearing in code — these are not wake requests. Activate only when the user opens with a wake word and clearly intends Archon to take over.