scale-colony
О программе
Этот навык предлагает стратегии масштабирования распределенных систем и команд, моделируя их по образу биологических колоний с использованием таких методов, как почкование и разделение ролей. Он помогает распознавать фазы роста, внедрять протоколы деления и выявлять пределы масштабирования, когда накладные расходы на коммуникацию начинают подавлять продуктивность. Используйте его превентивно, когда системы, работавшие с 10 агентами, перестают функционировать при 50, или при возникновении сбоев координации, таких как дублирование работы.
Быстрая установка
Claude Code
Рекомендуетсяnpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/scale-colonyСкопируйте и вставьте эту команду в Claude Code для установки этого навыка
Документация
Scale Colony
Scale distributed systems, teams, orgs through colony budding (split), role differentiation (age polyethism), and growth-triggered architectural transitions — keep coordination quality as colony grows past initial design.
When Use
- Team or system that worked at 10 agents breaks down at 50
- Comms overhead grows faster than productive output
- Coordination patterns implicit must become explicit
- Planning growth phase, want scale proactive not reactive
- Coordination failures correlate with size (lost messages, duplicated work, unclear ownership)
- Existing system needs split into semi-autonomous sub-colonies
Inputs
- Required: Current colony size + target growth (or growth rate)
- Required: Current coordination mechanisms + stress points
- Optional: Colony structure (flat, hierarchical, clustered)
- Optional: Role differentiation already in place
- Optional: Growth timeline + constraints
- Optional: Inter-colony coordination needs (if splitting)
Steps
Step 1: Recognize Growth Phase
Identify which scaling phase colony is in to apply right strategies.
- Classify current growth phase.
Colony Growth Phases:
┌───────────┬──────────────┬───────────────────────────────────────────┐
│ Phase │ Size Range │ Characteristics │
├───────────┼──────────────┼───────────────────────────────────────────┤
│ Founding │ 1-7 agents │ Everyone does everything, direct comms, │
│ │ │ implicit coordination, high agility │
├───────────┼──────────────┼───────────────────────────────────────────┤
│ Growth │ 8-30 agents │ Roles emerge, some specialization, comms │
│ │ │ overhead increases, need for structure │
├───────────┼──────────────┼───────────────────────────────────────────┤
│ Maturity │ 30-100 agents│ Formal roles, layered coordination, │
│ │ │ sub-groups form, inter-group coordination │
├───────────┼──────────────┼───────────────────────────────────────────┤
│ Fission │ 100+ agents │ Colony too large for single coordination │
│ │ │ framework, must bud into sub-colonies │
└───────────┴──────────────┴───────────────────────────────────────────┘
- Identify growth stress signals.
- Communication overload: messages per agent per day growing faster than colony size
- Decision latency: time from proposal to decision growing
- Coordination failures: duplicated work, dropped tasks, conflicting actions growing
- Knowledge dilution: new agents take longer to become productive
- Identity loss: agents cannot describe colony's purpose consistent
- Determine if colony is about to cross phase boundary or has crossed it
Got: Clear identification of current growth phase + specific stress signals showing colony approaching or crossed phase boundary.
If fail: Phase not clear? Measure three concrete metrics: comm volume per agent, decision latency, coordination failure rate. Plot over time. Inflection points = phase transitions. No metrics? Colony likely Founding phase (where metrics not yet needed).
Step 2: Implement Role Differentiation (Age Polyethism)
Introduce progressive specialization where agents take different roles by experience + colony needs.
- Define role progression path.
- Newcomers: observation, learning, simple tasks (low autonomy, high guidance)
- Workers: standard task execution, signal following (moderate autonomy)
- Specialists: domain expertise, complex tasks, mentor newcomers (high autonomy)
- Foragers/Scouts: exploration, innovation, external interface (see
forage-resources) - Coordinators: inter-group communication, conflict resolution, quorum management
- Implement role transitions.
- Triggered by experience thresholds, not appointment
- Agent that completed threshold of tasks transitions to next role (calibrate by complexity, growth — 5-10 for simple, 20-30 for specialist)
- Reverse transitions possible (specialist → worker in new domain)
- Colony's role distribution adapts to needs.
- Growing colony → more newcomer slots, active mentoring
- Stable colony → balanced distribution across all roles
- Threatened colony → more defenders, fewer scouts (see
defend-colony)
- Preserve role flexibility.
- No agent permanently locked into role
- Emergency protocols can temp reassign any agent to any role
- Cross-training ensures agents can cover adjacent roles
Got: Role structure where agents progress from simple to complex responsibilities, colony's role distribution reflects current needs + phase.
If fail: Role differentiation creates rigid silos? Up cross-training + rotation frequency. Newcomers struggle to progress? Mentoring insufficient — pair each newcomer with specialist for first N tasks. Too many cluster in one role? Transition triggers miscalibrated — adjust by colony-wide role demand.
Step 3: Restructure Coordination for Scale
Adapt coordination from coordinate-swarm to handle bigger colony.
- Replace direct comms with layered signaling.
- Founding: everyone talks to everyone (N×N comms)
- Growth: cluster into squads of 5-8; direct comms within squads, signal-based between
- Maturity: squads form departments; intra-squad direct, inter-squad signal, inter-department broadcast
- Implement coordination layers.
- Local coordination: within squad, direct signal exchange (stigmergy)
- Regional coordination: between squads in same department, aggregated signals
- Colony coordination: between departments, broadcast only for colony-wide decisions
- Design inter-layer interfaces.
- Each squad has one designated communicator who aggregates + relays signals
- Communicators filter noise: not every local signal relayed up
- Colony-wide broadcasts rare, reserved for quorum decisions, alarm escalation, major state changes
- Communication overhead budget.
- Target: each agent spends <20% of capacity on coordination
- Measure actual overhead; if exceeds budget, add another coordination layer or split oversized squad
Got: Layered coordination structure where comms overhead grows logarithmic (not linear) with colony size. Local coordination fast + direct; colony-wide slower but functional.
If fail: Coordination layers create info bottlenecks (communicators overloaded)? Add redundant communicators or reduce relay frequency. Layers create isolation (squads do not know what others do)? Up inter-layer signal frequency or create cross-squad liaison roles.
Step 4: Execute Colony Budding (Fission)
Split colony into semi-autonomous sub-colonies when exceeds single-coordination capacity.
- Recognize fission triggers.
- Colony exceeds 100 agents (or coordination layer count exceeds 3)
- Comms overhead exceeds 30% of agent capacity despite layering
- Decision latency exceeds acceptable thresholds for time-sensitive ops
- Subgroups have distinct identities, can operate independently
- Plan fission.
- Identify natural split lines (existing clusters, domain boundaries, geographic separation)
- Each daughter colony has viable role distribution (cannot split all specialists into one)
- Each daughter colony must have: at least one coordinator, sufficient workers, access to shared resources
- Define inter-colony interface: what info shared, what independent
- Execute split.
- Announce fission plan + timeline (consensus required — see
build-consensus) - Transfer agents to daughter colonies based on existing cluster membership
- Establish inter-colony comm channels (lightweight, async)
- Each daughter colony bootstraps own local coordination (inheriting patterns from parent)
- Announce fission plan + timeline (consensus required — see
- Post-fission stabilization.
- Monitor each daughter for viability (can sustain itself?)
- Inter-colony coordination minimal (quarterly sync, not daily)
- If daughter fails, reabsorb into nearest viable colony
Got: Two or more viable daughter colonies, each operating semi-autonomous with own coordination, connected by lightweight inter-colony interfaces.
If fail: Daughter colonies too small to be viable? Fission was premature — remerge, try again at larger size. Inter-colony coordination as heavy as pre-fission single-colony? Split lines wrong — colonies too interdependent. Re-draw boundaries along natural independence lines.
Step 5: Monitor Scaling Limits and Adapt
Continuous assess if current structure matches colony's size + needs.
- Track scaling health metrics.
- Coordination overhead ratio: time coordinating / time producing
- Decision throughput: decisions per time unit (should grow or hold steady with growth)
- Agent satisfaction: engagement, retention, sense of purpose (drops when scaling fails)
- Error rate: coordination failures per time unit (should not grow linear with growth)
- Identify scaling limit indicators.
- Overhead ratio exceeding 25% → need more automation or another coordination layer
- Decision throughput declining → governance structure needs revision
- Agent turnover spiking → cultural or structural issues from scaling
- Error rate accelerating → coordination mechanisms failing
- Trigger adaptation.
- Phase transition detected → apply right phase strategy from Step 1
- Scaling limit reached → escalate to next structural intervention (role differentiation → coordination restructure → fission)
- External change (market shift, tech disruption) → may require colony transformation (see
adapt-architecture)
Got: Colony that monitors own scaling health, proactively adapts structure before scaling stress becomes scaling failure.
If fail: Scaling health metrics not available? Colony lacks observability — build measurement before more structure. Metrics show problems but colony cannot adapt? Resistance cultural not technical — address human factors (fear of change, ownership attachment, trust deficits) before restructuring.
Checks
- Current growth phase identified with specific stress signals
- Role differentiation defined with progressive specialization
- Coordination layered for colony size
- Comms overhead stays below 20-25% of agent capacity
- Fission plan exists for when colony exceeds single-coordination capacity
- Scaling health metrics tracked, thresholds trigger adaptation
- Each daughter colony (post-fission) has viable role distribution
Pitfalls
- Scale structure before needed: Premature layering adds overhead without benefit. 10-person team does not need department coordinators. Let stress signals guide structural changes
- Preserve founding culture at all costs: What worked at 5 agents will not work at 50. Scaling needs structural evolution; nostalgia for founding phase prevents necessary adaptation
- Fission without independence: Splitting colony into sub-colonies that still depend on each other for daily ops = worst of both worlds — overhead of coordination + overhead of separation
- Uniform role distribution: Not every sub-colony needs same role ratios. Research colony needs more scouts; production needs more workers. Adapt role distribution to mission
- Ignore remerge as option: Sometimes fission fails, best move is remerge. Treat fission as irreversible = prevents recovery from bad splits
See Also
coordinate-swarm— foundational coordination patterns this skill scalesforage-resources— foraging scales different than production; role differentiation affects scout allocationbuild-consensus— consensus mechanisms must adapt for larger groupsdefend-colony— defense must scale with colonyadapt-architecture— morphic skill for structural transformation, triggered by growth pressureplan-capacity— capacity planning for growth projectionsconduct-retrospective— retrospectives help identify scaling stress before failure
GitHub репозиторий
Похожие навыки
railway-docs
ДокументацияЭтот навык получает актуальную документацию Railway, чтобы отвечать на вопросы о функциях, возможностях или конкретных URL-адресах документации. Он гарантирует, что разработчики получают точную и современную информацию напрямую из официальных источников Railway. Используйте его, когда пользователи спрашивают, как работает Railway, или ссылаются на документацию Railway.
n8n-code-python
ДокументацияЭтот навык Claude предоставляет экспертные рекомендации по написанию кода Python в узлах Code платформы n8n, в частности, по использованию стандартной библиотеки Python и работе со специальным синтаксисом n8n, таким как `_input`, `_json` и `_node`. Он помогает разработчикам понять ограничения Python в среде n8n и рекомендует использовать JavaScript для большинства рабочих процессов, предлагая решения на Python для конкретных задач по преобразованию данных.
archon
ДокументацияНавык Archon предоставляет семантический поиск на основе RAG и управление проектами через REST API. Используйте его для запросов к документации, управления иерархическими проектами/задачами и выполнения поиска информации с возможностью загрузки документов. Всегда в первую очередь обращайтесь к Archon при поиске во внешней документации, прежде чем использовать другие источники.
n8n-code-javascript
ДокументацияЭтот навык Claude предоставляет экспертные рекомендации по написанию кода JavaScript в узлах Code платформы n8n. Он охватывает важный синтаксис, специфичный для n8n, включая переменные `$input`/`$json`, HTTP-хелперы и работу с DateTime, а также помогает в устранении распространённых ошибок. Используйте его при разработке рабочих процессов в n8n, требующих кастомной обработки JavaScript в узлах Code.
