← ClaudeAtlas

fablelisted

Run a software-delivery request through selective routing with Claude Fable 5.1 as the primary orchestrator. The heavyweight chair for the hardest or highest-stakes work; the route floor is audit, so independent review is the norm rather than the exception. Use when stakes exceed what the everyday opus chair should own.
toshon-jennings/agent-advisor · ★ 0 · AI & Automation · score 55
Install: claude install-skill toshon-jennings/agent-advisor
# Fable chair Treat this skill's arguments as the user's task request. This is the heavyweight chair. Stakes decide the chair, not subject matter. Establish these three values, then follow the authoritative workflow: ~~~text chair: Claude Fable 5.1 floor: audit failover: claude-advisor:reviewer ~~~ Before taking any task action, read [`../orchestration/SKILL.md`](../orchestration/SKILL.md) completely and follow it as the authoritative workflow, along with any reference it requires. Reading it is a setup action; do not inspect or modify the user's workspace until that workflow has declared its `SELECTIVE ROUTE`. If the primary session is not Fable 5.1, stop and give the user two options: switch the chair to Fable 5.1, or drop to the everyday `opus` chair and accept its lower floor. Never proceed at this chair from another model. **Why the floor is audit.** Entering this chair is itself a recorded claim that the stakes are high, so independent review is the floor rather than the exception. `full` is ordinary here. `delegate` drops review — take it only when the residual risk is in the building, not the judging, and record why. `solo` is below the floor: it needs a recorded justification, and if it is genuinely right, the everyday chair was probably the correct chair. **Review independence at this chair.** Review defaults to the out-of-process `reviewer-codex` lane, which is cross-vendor. This chair's in-process failover is `claude-advisor:reviewer`, pinned to Opus while