research-runlisted

# research-run ## Purpose `/research-run` executes an existing Package one Experiment at a time. It starts only after Project, Direction, Experiment, and Package records exist in management state. If one is missing, the command returns a handoff to the skill that owns it. The command keeps running or monitoring until it reaches one of these outcomes: - the Package has an evidence-backed terminal result; - a proposal or user decision blocks further work; - an external dependency has a concrete blocker and a recorded next action. It never invokes git. ## Authority boundary Management state lives in `.research/state/events.jsonl`. Run measurements live under `.research/experiments/`. Files under `.research/interface/` are generated views for people. `research-run` does not read them, and a missing interface does not block execution. `research-run` may: - query the selected Package, Experiment, open Runs, Decisions, Rules, and evidence; - dispatch implementation, review, launch, monitoring, and verification work; - launch or inspect a Run through `lib.experiments`; - emit commands for `research-op`; - reconcile experiment callbacks with `research-op --op scan-events`. It may not: - create or revise Project, Direction, or Experiment intent; - materialize a Package; - append management events directly; - treat generated HTML or JavaScript as state; - copy a measured value into an Experiment gate. ## Code entry points | Purpose | Entry point | | --- | --- | | Admission