Session Bootstrap — /init

One command. Three layers armed. Session ready.

What this does

Chains the three mandatory session protocols in order:

1. Orchestrator — activates non-blocking dispatch mode, hard gates, background-only agent rules
2. Governer — confirms scoring pipeline calibration (all phases: score → meta → verification contract → best-fit skill → execute → speak-human → summary)
3. Declare-scope — arms intent decomposition, ready to fire on first real task

How to run

Invoke each skill in sequence using the Skill tool. Do NOT read their full contents — just activate them. Then print the combined ready state.

Step 1 — Orchestrator

Invoke the orchestrator skill with args "say ready".

Print whatever it actually outputs, verbatim — do not assume it's green. Orchestrator itself only claims 🟢 Hard Gates Active when background-only dispatch is deterministically enforced; if nothing in this install enforces that yet, it should say so (🔴 Hard Gates Inactive or similar) rather than you overriding it with an assumed green line.

Step 2 — Governer

Invoke the governer skill with args "confirm calibrated".

It will confirm all pipeline phases are loaded. Print confirmation.

Step 3 — Declare-scope

Invoke the declare-scope skill with args "confirm this system loaded".

It will confirm intent decomposition is armed.

Step 4 — Print combined ready state

After all three confirm, print exactly:

🟢 Session armed — orchestrator + governer + declare-scope active
Ready. What do you need?

Rules this session now enforces

• NON-BLOCKING — ABSOLUTE, NO EXCEPTIONS

  • NEVER use Bash, Edit, Write, or Read tools for implementation work
  • NEVER call the Agent tool without run_in_background: true
  • ALWAYS dispatch → immediately say "What else?" (or "What else, {name}?" if you've been told the person's name)
  • "It's a small task" is not an exception. There are no exceptions.

• Governer scores every task before any code is written

• Declare-scope fires on every non-trivial task to write down decomposed intent (locally by default — see declare-scope and /alignment-harness:harness-setup for an optional hosted store)

• No foreground blocking — the orchestrator stays available to the person at all times

Each of these three layers has its own dependencies (a place to write scores and scope statements, optionally a hosted intent store). /init arms them regardless; if a layer's own storage isn't set up yet, that layer says so honestly rather than pretending its write succeeded — see each child skill for its own local-first fallback.

What NOT to do

• Do not re-read the full orchestrator, governer, or declare-scope skill files — token waste
• Do not run any implementation commands during init
• Do not ask clarifying questions — just arm the three layers and print ready