adapt-physarum-router
Domain: adapt · Model class: cheap
Description
Section titled “Description”Use when a user wants workflow routing topology to self-prune — automatically removing underperforming paths and reinforcing high-throughput ones. Triggers: “self-pruning workflow”, “slime mould optimisation”, “Physarum”, “reinforce busy paths”, “prune unused routes”, “adaptive network topology”, “workflow that removes dead ends”. Best for flows with 6+ edges. Also trigger when a user notices some workflow paths are never used and wants the system to consolidate onto paths that matter.
Purpose
Section titled “Purpose”Model each edge as tube with conductance D (init 1.0). After each cycle: D(t+1)=D(t)×|flow(t)|^μ (normalised). Prune edges where D<pruning_threshold. Spawn exploratory edges with p_explore.
Trigger Phrases
Section titled “Trigger Phrases”- “self-pruning workflow”
- “slime mould optimisation”
- “Physarum”
- “reinforce busy paths”
- “prune unused routes”
- “adaptive network topology”
- “workflow that removes dead ends”
- “workflow paths are never used”
Anti-Triggers
Section titled “Anti-Triggers”- the user wants a one-off improvement without ongoing adaptation or structural change
Intake Questions
Section titled “Intake Questions”- What initial edges are eligible for reinforcement or pruning?
- How is throughput or flow measured on each edge?
- What decay, reinforcement, and prune thresholds fit the workload?
- How many adaptation cycles should run before edges can be removed?
Output Contract
Section titled “Output Contract”- routing decision artifact
- configuration and telemetry summary
- next-action explanation
- validation or operator notes
Related Skills
Section titled “Related Skills”flow-orchestrator · orch-agent-orchestrator · prompt-chaining