--- description: Enable NERVous System workflow for this task argument-hint: "[drain=] [risk=] [task]" --- Use the NERVous System for this invocation. Invocation arguments: $ARGUMENTS Optional invocation configuration tokens may be included in the arguments: `drain=`, `risk=`, `policy=`, `evidence="..."`, `dangerous_opt_in=true`. If present, first call `cortex set_config` with those values, then treat the remaining arguments as the task. If `risk=disabled` is requested without explicit `dangerous_opt_in=true` and non-empty `evidence`, stop and ask the user for confirmation/evidence instead of proceeding. Model defaults for LION/MAGI are persisted separately with `/nervous:config lion_model=... lion_fallback_model=... magi_model=... magi_fallback_model=...` before invoking this prompt; unset model defaults preserve pi's current/default model behavior. Default mode: drain the active NERVous context. Keep progressing until every workable incomplete CORTEX goal in this context is completed. Workable includes normal actionable goals, skipped/blocked goals due for revisit, and failed goals due for retry or retryability classification. If a goal cannot safely proceed, escalate it to AMYGDALA or mark it cancelled/blocked with clear evidence plus revisit/retry metadata; do not silently abandon it. - Start by checking CORTEX config and applying invocation config tokens if supplied; users can also persist defaults with `/nervous:config` before invoking this prompt. - Check CORTEX for existing incomplete goals before creating new unrelated goals. - If this prompt introduces new work, perform one bounded task-framing pass before `cortex analyze`: inspect relevant project context when useful; make the objective, scope, non-goals, assumptions, success criteria, candidate options, and any MAGI decision concrete. Ask the user only about ambiguity that blocks a responsible goal; record non-blocking uncertainty as assumptions or open questions. Persist the result in `cortex analyze.framing`. If analyze reports MAGI readiness gaps, repair that same analyzed goal with `cortex refine` rather than creating a duplicate, then include it in the same drain loop. - Framing happens once for new work. Do not repeat it when resuming, replanning, retrying, or revisiting an existing goal; use the durable framing already stored on the goal. - Check CORTEX config/risk_gate_mode, then drain/list goals; for each workable CORTEX goal: set/resume it, inspect linked AXON tasks, execute ready/retry/revisit work, verify against success criteria, and complete it when verified. - Persist durable subtasks/status in AXON. - Use SYNAPSE only for short coordination/risk notes. - Use GANGLION to allocate capable LION workers, and CEREBEL to orchestrate/record worker waves. - Delegate narrow implementation or review work to LION when it reduces risk or context load; use `model_role="review"` for LION review/QA assignments. - Escalate blockers, security/data-loss/regression risk, or unsafe uncertainty to AMYGDALA unless risk_gate_mode has explicit approval evidence (`auto_deliberate`, `user_accepted`, or dangerous `disabled`). - Use MAGI for hard, ambiguous, risky, or architecturally significant decisions before planning. Before convening, ensure the framed objective, scope, success criteria, and `decision_needed` are concrete; map framing context, constraints, candidate options, and decision needed into the MAGI call. Consider MAGI final review for high-impact completed goals. - Record failed work with retryability; record skipped work with next_revisit_at/unblock conditions; revisit due skipped work and use `cortex reopen` when resolved so it can be replanned/completed. - Repeat until no workable CORTEX goals remain. Final answer: completed goals, waiting/retry/skipped/cancelled goals, and evidence. - Keep the answer concise: component → action/status → evidence.