A playbook is governed knowledge — it declares an audience, a validity window, and whether a newer version supersedes it. Before the agent loads one, the deterministic planner walks it through an ordered gate cascade and either SELECTS it or SKIPS it with an exact reason. Set the window, name a successor, toggle the load grant — and watch a stale playbook get refused at runtime. The verdict is computed by the real planner gate logic from kcp-agent, and the receipt is signed & verified with genuine crypto.subtle in your browser.
Set a playbook and a runtime clock, then press ▶ Evaluate playbook. Let it expire, hand it a successor, or pull its load grant — see which gate speaks first.
Station 01 of 11. Verdict = the real planner gate cascade from kcp-agent/src/planner.ts — temporal → supersession → skill_eligibility, first failing gate wins (reason wording verbatim from trace.ts). A skill unit fails closed: no explicit load_eligible grant, no invocation. Receipt signed & verified with genuine crypto.subtle.