// Single source of truth protocol for Wild V2 preflight, sweep/run auditability, GPU discovery, and parallel scheduling
System prompt for handling experiment alerts. Provides diagnosis guidance, GPU wrapper context, action suggestions, and structured response from allowed choices.
Default system prompt for agent chat mode. Provides identity, environment context, compute awareness, API-driven job submission, and workflow reflection.
Generates a structured experiment plan with compute-aware recommendations and saves it via the plan API endpoint.
Wraps user steering input with context signals for the model during a wild loop session
Ports new models into FastVideo with strict numerical alignment to official implementations. Use when adding a FastVideo model/pipeline, porting an official or Diffusers checkpoint, or debugging parity/alignment.
Protocol for GPU discovery and parallel run scheduling across local GPU and Slurm clusters
| name | wild_v2_execution_ops_protocol |
| description | Single source of truth protocol for Wild V2 preflight, sweep/run auditability, GPU discovery, and parallel scheduling |
| category | protocol |
| variables | [] |
This is the canonical protocol for execution operations in Wild V2. Use this skill as the source of truth instead of duplicating long operational instructions in prompts.
Before planning or execution, verify:
curl -sf "$SERVER_URL/docs" >/dev/null
curl -sf "$SERVER_URL/openapi.json" >/dev/null
curl -sf "$SERVER_URL/prompt-skills/wild_v2_execution_ops_protocol" >/dev/null
curl -sf "$SERVER_URL/wild/v2/system-health" >/dev/null
If preflight fails, abort loop work immediately.
POST /runs.sweep_id) created via POST /sweeps/wild.POST /runs, POST /sweeps, POST /sweeps/wild, POST /sweeps/{id}/runs), include chat_session_id in the JSON body using the current chat session id.POST /runs), one per config.curl -X POST "$SERVER_URL/cluster/detect" -H "X-Auth-Token: $AUTH_TOKEN"
curl -X GET "$SERVER_URL/cluster" -H "X-Auth-Token: $AUTH_TOKEN"
curl -X GET "$SERVER_URL/wild/v2/system-health" -H "X-Auth-Token: $AUTH_TOKEN"
Use:
cluster.typecluster.gpu_countsystem-health.runningLet:
g = max(1, cluster.gpu_count or 1) for local_gpug = max(1, cluster.gpu_count or 4) for slurm (conservative default if unknown)g = 1 for cpu_only or unknownr = system_health.runningq = number_of_ready_or_queued_runsThen:
target_parallel = g
max_new_runs = max(0, min(q, target_parallel - r))
Interpretation:
max_new_runs additional runs now.max_new_runs == 0, wait or stop low-priority runs before starting more.CUDA_VISIBLE_DEVICES=i).max_new_runs.auto_start=true for up to max_new_runs, orPOST /runs/{id}/start for selected runs.GET /runs and GET /wild/v2/system-health.