From agent-almanac
Coordinates distributed agents via stigmergy, local rules, and quorum sensing to achieve emergent collective behavior without centralized control. Use for scaling multi-agent systems, self-organizing teams, or event-driven architectures.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agent-almanac:coordinate-swarmThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Establish coordination across distributed agents using stigmergy (indirect communication through environment modification), local interaction rules, and quorum sensing — enabling coherent collective behavior without a central controller.
Establish coordination across distributed agents using stigmergy (indirect communication through environment modification), local interaction rules, and quorum sensing — enabling coherent collective behavior without a central controller.
forage-resources, build-consensus)Classify the coordination challenge to select appropriate patterns.
forage-resources)build-consensus)defend-colony)Expected: A clear classification of the coordination problem type and the specific failure mode to address. This determines which swarm patterns to apply.
On failure: If the problem doesn't fit a single class, it may be a composite. Decompose into sub-problems and address each with the appropriate pattern. If agents are too heterogeneous for a single coordination model, consider layered coordination — homogeneous clusters coordinated via inter-cluster stigmergy.
Create the indirect communication channels through which agents influence each other's behavior.
Signal Design Template:
┌──────────────┬───────────────────┬──────────────┬────────────────────┐
│ Signal Name │ Deposited When │ Decay Rate │ Agent Response │
├──────────────┼───────────────────┼──────────────┼────────────────────┤
│ success-trail│ Task completed OK │ 50% per hour │ Follow toward │
│ busy-marker │ Agent starts task │ On completion│ Avoid / pick other │
│ help-signal │ Agent stuck >5min │ 25% per hour │ Assist if nearby │
│ danger-flag │ Error detected │ 10% per hour │ Retreat & report │
└──────────────┴───────────────────┴──────────────┴────────────────────┘
Expected: A signal table mapping environmental markers to agent deposit conditions, decay rates, and response behaviors. Signals should be simple, composable, and independently meaningful.
On failure: If signal design feels overly complex, reduce to two signals: one positive (success trail) and one negative (danger flag). Most coordination problems can be bootstrapped with attract/repel dynamics. Add nuance only after the basic system is functioning.
Specify the simple rules each agent follows, using only local information (their own state + nearby signals).
Expected: A prioritized rule set that each agent executes independently. When applied across the swarm, these local rules produce the target collective behavior (foraging, construction, defense, etc.).
On failure: If mental simulation doesn't produce the desired emergent behavior, the rules likely need a feedback loop — agents must be able to observe the consequences of their collective actions. Add a signal that represents the collective state (e.g., "task completion rate") and a rule that adjusts behavior based on it.
Set thresholds that trigger collective state changes when enough agents agree.
Expected: Quorum thresholds that allow the swarm to make collective decisions without a leader. The hysteresis gap prevents rapid oscillation between states.
On failure: If the swarm oscillates between states, widen the hysteresis gap (e.g., activate at 70%, deactivate at 30%). If the swarm never reaches quorum, lower the threshold or increase the sensing window. If decisions are too slow, reduce the sensing window — but beware of premature consensus.
Validate that local rules produce the desired collective behavior, then tune parameters.
Expected: A tuned parameter set where the swarm self-organizes toward the target behavior, recovers from perturbations, and scales gracefully.
On failure: If the swarm fails stress tests, the signal design is likely too tightly coupled. Simplify: reduce to fewer signals, increase decay rates (fresher information), and ensure agents have a robust default behavior when no signals are present. A swarm that does something reasonable with zero signals is more resilient than one that depends on signal availability.
scale-colony)forage-resources — applies swarm coordination specifically to resource search and explore-exploit tradeoffsbuild-consensus — deep dive into distributed agreement mechanisms, extending the quorum sensing from this skilldefend-colony — collective defense patterns that build on the signal and rule framework herescale-colony — scaling strategies for when the swarm outgrows its initial coordination designadapt-architecture — morphic skill for transforming system architecture, complementary when swarm coordination triggers structural changedeploy-to-kubernetes — practical distributed system deployment where swarm coordination patterns applyplan-capacity — capacity planning informed by swarm scaling dynamicscoordinate-reasoning — AI self-application variant; maps stigmergic signals to context management with information decay rates and local protocolsnpx claudepluginhub pjt222/agent-almanacScales distributed systems and organizations using colony budding, role differentiation, and architectural transitions to maintain coordination as agent counts grow past initial capacity.
Coordinates multi-agent swarms with queen-led architecture, Byzantine consensus, and shared persistent memory. Use for orchestrating agent teams under a lead coordinator or building resilient cross-agent collaboration.
Selects optimal subagent topology (hierarchical, mesh, ring, star) based on task structure and coordinates multi-agent swarm execution for complex, multi-file coding tasks.