FIXTURE—IDLEEXEC · —WS idleverifier v0

Agent Studio

01 Identity & archetype

LLM-Driftarena contestant · no Studio deployLLM-Drift: pinned fair-comparison arena contestant — the head-to-head opponent for det-Drift in the Arena, never a Studio deploy (the generic llm agent is not LLM-Drift).Use in Arena →

Source mode data source — replay or live.: STUDIO (reproducible). Proof mode reproducible = deterministic strategy rerun regenerates actions/scores; verified = recorded action/evidence recomputed and checked, but LLM/runtime not byte-reproduced; partial = shown for transparency, not fully eligible/complete.: reproducible. BYOA agents are verified, not reproducible (Phase-3).

02 Decision Shell (LLM)

Not applicable in this mode.

03 Deterministic config

Numeric thresholds: min_edge_bps, conviction, quote-age.

min_edge_bps8
max_quote_age_s30

04 Market scope

1X2_PARTICIPANT_RESULTOVERUNDER_PARTICIPANT_GOALS

05 Policy envelope & execution

max_stake capped policy sizing; never rank/scoring/proof.100
kill_switchfalse

Reviewable changes

Edits are a structured patch, applied as a new pinned version on PIN — never a live mutation of a running agent.

No pending changes.