UniqueSoft/APAW
2
0
Fork 0
Code Issues 15 Pull Requests Actions Packages Projects Releases Wiki Activity

[Research] Multi-Agent System Patterns - Anthropic and Kilo.ai #25

New Issue
Closed
opened 2026-04-05 00:49:56 +00:00 by NW · 4 comments
NW commented 2026-04-05 00:49:56 +00:00
Owner
Copy Link

Overview

Research milestone for multi-agent system improvements. Key findings from Anthropics Building Effective Agents and Kilo.ai documentation.

Key Findings

Anthropic - Building Effective Agents

Core Patterns:

  1. Prompt Chaining - Sequential steps with validation
  2. Routing - Classify and route to specialized agent
  3. Parallelization - Run independent tasks simultaneously
  4. Orchestrator-Workers - Central coordinator with specialized workers
  5. Evaluator-Optimizer - Iterative improvement loop

Best Practices:

  • Most successful implementations use simple, composable patterns
  • Avoid complex frameworks when simple patterns suffice
  • Trade-off: latency and cost for better task performance
  • Workflows = predictability, Agents = flexibility

Kilo.ai - Custom Subagents

Key Concepts:

  • mode: subagent for isolated context
  • mode: primary for main agent
  • mode: all for full tool access
  • Permission system: allow, ask, deny per tool

Applied to APAW

Pattern Implementation Status
Routing orchestrator.md Done
Orchestrator-Workers workflow.md Done
Evaluator-Optimizer code-skeptic loop Needs formalization
Parallelization review phase Not implemented

References

Related Issues

  • Issue #21 - Implement parallelization
  • Issue #22 - Formalize evaluator-optimizer
  • Issue #23 - Add quality gate enforcement
## Overview Research milestone for multi-agent system improvements. Key findings from Anthropics Building Effective Agents and Kilo.ai documentation. ## Key Findings ### Anthropic - Building Effective Agents **Core Patterns:** 1. **Prompt Chaining** - Sequential steps with validation 2. **Routing** - Classify and route to specialized agent 3. **Parallelization** - Run independent tasks simultaneously 4. **Orchestrator-Workers** - Central coordinator with specialized workers 5. **Evaluator-Optimizer** - Iterative improvement loop **Best Practices:** - Most successful implementations use simple, composable patterns - Avoid complex frameworks when simple patterns suffice - Trade-off: latency and cost for better task performance - Workflows = predictability, Agents = flexibility ### Kilo.ai - Custom Subagents **Key Concepts:** - mode: subagent for isolated context - mode: primary for main agent - mode: all for full tool access - Permission system: allow, ask, deny per tool ## Applied to APAW | Pattern | Implementation | Status | |--------|---------------|--------| | Routing | orchestrator.md | Done | | Orchestrator-Workers | workflow.md | Done | | Evaluator-Optimizer | code-skeptic loop | Needs formalization | | Parallelization | review phase | Not implemented | ## References - https://www.anthropic.com/research/building-effective-agents - https://docs.kilo.ai/custom-subagents ## Related Issues - Issue #21 - Implement parallelization - Issue #22 - Formalize evaluator-optimizer - Issue #23 - Add quality gate enforcement
NW added the type::documentationtype::chore labels 2026-04-05 00:49:56 +00:00
NW commented 2026-04-05 01:01:51 +00:00
Author
Owner
Copy Link

? Cognitive Enhancement Agents Implemented

New Agents (Based on Research)

Agent Pattern Source
@planner CoT, ToT, Plan-Execute-Reflect Lilian Weng, Anthropic
@reflector Reflexion Pattern Shinn & Labash 2023
@memory-manager Memory Systems Lilian Weng 2023

New Skills

Skill Description Source
memory-systems Short/Long/Episodic memory Lilian Weng
planning-patterns CoT, ToT, ReAct, Reflexion Multiple
tool-use ACI Design Principles Anthropic

New Rules

Rule Description
agent-patterns Core patterns from Anthropic, OpenAI, Lilian Weng

Key Research Findings Applied

Anthropic - Building Effective Agents:

  • ? Prompt Chaining - Sequential steps with validation
  • ? Routing - Classify and route to specialized agent
  • ? Parallelization - Run independent tasks (Issue #21)
  • ? Orchestrator-Workers - Central controller with workers
  • ? Evaluator-Optimizer - Iterative improvement (Issue #22)

Lilian Weng - LLM Powered Autonomous Agents:

  • ? Planning: Task decomposition, self-reflection
  • ? Memory: Short-term, long-term, episodic
  • ? Tool Use: External APIs, code execution

Commit: 774dc9a

## ? Cognitive Enhancement Agents Implemented ### New Agents (Based on Research) | Agent | Pattern | Source | |-------|---------|--------| | `@planner` | CoT, ToT, Plan-Execute-Reflect | Lilian Weng, Anthropic | | `@reflector` | Reflexion Pattern | Shinn & Labash 2023 | | `@memory-manager` | Memory Systems | Lilian Weng 2023 | ### New Skills | Skill | Description | Source | |-------|-------------|--------| | `memory-systems` | Short/Long/Episodic memory | Lilian Weng | | `planning-patterns` | CoT, ToT, ReAct, Reflexion | Multiple | | `tool-use` | ACI Design Principles | Anthropic | ### New Rules | Rule | Description | |------|-------------| | `agent-patterns` | Core patterns from Anthropic, OpenAI, Lilian Weng | ### Key Research Findings Applied **Anthropic - Building Effective Agents:** - ? Prompt Chaining - Sequential steps with validation - ? Routing - Classify and route to specialized agent - ? Parallelization - Run independent tasks (Issue #21) - ? Orchestrator-Workers - Central controller with workers - ? Evaluator-Optimizer - Iterative improvement (Issue #22) **Lilian Weng - LLM Powered Autonomous Agents:** - ? Planning: Task decomposition, self-reflection - ? Memory: Short-term, long-term, episodic - ? Tool Use: External APIs, code execution **Commit:** 774dc9a
NW commented 2026-04-05 01:07:29 +00:00
Author
Owner
Copy Link

? Research Cycle Implemented

Continuous Self-Improvement Added

New Skill: research-cycle

  • Monitors agent performance
  • Identifies capability gaps
  • Researches best practices
  • Implements improvements
  • Tracks evolution in Gitea

New Command: /research

  • Manual trigger for research
  • Fetches from Anthropic, OpenAI, Lilian Weng
  • Creates agents/skills based on findings

Commits

Milestone: #47

## ? Research Cycle Implemented ### Continuous Self-Improvement Added New Skill: research-cycle - Monitors agent performance - Identifies capability gaps - Researches best practices - Implements improvements - Tracks evolution in Gitea New Command: /research - Manual trigger for research - Fetches from Anthropic, OpenAI, Lilian Weng - Creates agents/skills based on findings ### Commits - 774dc9a: Cognitive agents - 420c10f: Capability integration - 348c47f: Research cycle Milestone: #47
NW commented 2026-04-05 01:10:49 +00:00
Author
Owner
Copy Link

? Research Milestone Complete

Implemented Improvements

Issue Pattern Status
#21 Parallelization ? Closed - workflow template created
#22 Evaluator-Optimizer ? Closed - iteration_loops in capability-index
#23 Quality Gates ? Closed - quality_gates configured
#24 Agent Consolidation ? Closed - deferred, capability routing added

New Cognitive Agents

Agent Pattern Source
@planner CoT, ToT, Plan-Execute-Reflect Lilian Weng, Anthropic
@reflector Reflexion Shinn & Labash 2023
@memory-manager Memory Systems Lilian Weng 2023

New Skills

Skill Description
memory-systems Short/Long/Episodic memory
planning-patterns CoT, ToT, ReAct, Reflexion
tool-use ACI Design Principles
research-cycle Continuous self-improvement

New Workflows

Workflow Pattern
parallel-review Parallel execution
evaluator-optimizer Iterative improvement
chain-of-thought Sequential decomposition

Files Modified

  • .kilo/agents/planner.md - NEW
  • .kilo/agents/reflector.md - NEW
  • .kilo/agents/memory-manager.md - NEW
  • .kilo/capability-index.yaml - Updated with cognitive agents
  • .kilo/agents/orchestrator.md - Added routing for new agents
  • AGENTS.md - Updated agent list
  • .kilo/workflows/*.md - NEW (3 templates)
  • .kilo/skills/*/SKILL.md - NEW (4 skills)

Milestone

  • Created: Milestone #47 - Cognitive Enhancement

Commits

All research findings applied. Continuous research cycle active.

## ? Research Milestone Complete ### Implemented Improvements | Issue | Pattern | Status | |-------|---------|--------| | #21 | Parallelization | ? Closed - workflow template created | | #22 | Evaluator-Optimizer | ? Closed - iteration_loops in capability-index | | #23 | Quality Gates | ? Closed - quality_gates configured | | #24 | Agent Consolidation | ? Closed - deferred, capability routing added | ### New Cognitive Agents | Agent | Pattern | Source | |--------|---------|--------| | @planner | CoT, ToT, Plan-Execute-Reflect | Lilian Weng, Anthropic | | @reflector | Reflexion | Shinn & Labash 2023 | | @memory-manager | Memory Systems | Lilian Weng 2023 | ### New Skills | Skill | Description | |-------|-------------| | memory-systems | Short/Long/Episodic memory | | planning-patterns | CoT, ToT, ReAct, Reflexion | | tool-use | ACI Design Principles | | research-cycle | Continuous self-improvement | ### New Workflows | Workflow | Pattern | |----------|--------| | parallel-review | Parallel execution | | evaluator-optimizer | Iterative improvement | | chain-of-thought | Sequential decomposition | ### Files Modified - `.kilo/agents/planner.md` - NEW - `.kilo/agents/reflector.md` - NEW - `.kilo/agents/memory-manager.md` - NEW - `.kilo/capability-index.yaml` - Updated with cognitive agents - `.kilo/agents/orchestrator.md` - Added routing for new agents - `AGENTS.md` - Updated agent list - `.kilo/workflows/*.md` - NEW (3 templates) - `.kilo/skills/*/SKILL.md` - NEW (4 skills) ### Milestone - Created: Milestone #47 - Cognitive Enhancement ### Commits - 774dc9a: Add cognitive agents - 420c10f: Integrate into capability index - 348c47f: Add research cycle - fbc1f61: Add workflow templates --- All research findings applied. Continuous research cycle active.
NW commented 2026-04-05 23:25:31 +00:00
Author
Owner
Copy Link

📋 Status Update

Completed

  • ✅ Routing pattern (orchestrator.md)
  • ✅ Orchestrator-Workers (workflow.md)

Research Findings Applied

Patterns documented in .kilo/rules/agent-patterns.md:

  • Prompt Chaining
  • Routing
  • Parallelization
  • Orchestrator-Workers
  • Evaluator-Optimizer

Notes

Research milestone serves as reference for future agent improvements. Closing as research phase complete.

## 📋 Status Update ### Completed - ✅ Routing pattern (orchestrator.md) - ✅ Orchestrator-Workers (workflow.md) ### Research Findings Applied Patterns documented in `.kilo/rules/agent-patterns.md`: - Prompt Chaining - Routing - Parallelization - Orchestrator-Workers - Evaluator-Optimizer ### Notes Research milestone serves as reference for future agent improvements. Closing as research phase complete.
NW closed this issue 2026-04-05 23:25:32 +00:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
agent::agent-architect
Owned by agent-architect
 agent::capability-analyst
Owned by capability-analyst
 agent::code-skeptic
Owned by code-skeptic
 agent::evaluator
Owned by evaluator
 agent::history-miner
Owned by history-miner
 agent::lead-developer
Owned by lead-developer
 agent::orchestrator
Owned by orchestrator
 agent::sdet-engineer
Owned by sdet-engineer
 agent::system-analyst
Owned by system-analyst
 agent::the-fixer
Owned by the-fixer
 budget::exhausted
Token budget exhausted
 budget::sufficient
Token budget sufficient
 budget::warning
Token budget low
 cascade::depth-0
No subagent calls
 cascade::depth-1
1-level subagent calls
 cascade::depth-2
2-level subagent calls
 cascade::depth-exceeded
Depth limit exceeded
 cascade::depth-n
Unlimited subagent calls
 dashboard::frozen
Dashboard feature freeze — no new features until thaw
 evolution::model-change
Model change evolution
 evolution::new-agent
New agent evolution
 evolution::new-skill
New skill evolution
 evolution::new-workflow
New workflow evolution
 evolution::prompt-opt
Prompt optimization evolution
 memory::checkpoint
Checkpoint stored
 memory::fresh
Checkpoint fresh
 memory::recoverable
Checkpoint recoverable
 memory::stale
Checkpoint stale
 permission::evolve-system
Can evolve system
 permission::read-only
Read-only access
 permission::violation
Security violation
 permission::write-code
Can write code
 permission::write-config
Can write config
 phase::awaiting-review
Agent awaits review
 phase::drafting-spec
Agent is drafting specification
 phase::executing
Agent is executing task
 phase::gathering-evidence
Agent is gathering data
 phase::refining-prompt
Agent is refining prompts
 phase::researching
Research phase — gathering evidence and solutions
 phase::verifying
Agent is verifying results
 priority::critical
Critical priority
 priority::high
High priority
 priority::low
Low priority
 priority::medium
Medium priority
 quality::blocked
Blocked by quality
 quality::fail
Quality check failed
 quality::needs-fix
Needs fixes
 quality::pass
Quality check passed
 size::l
Large (4-8 hours)
 size::m
Medium (2-4 hours)
 size::s
Small (1-2 hours)
 size::xl
Extra large (>8 hours)
 size::xs
Extra small (<1 hour)
 status::blocked
Blocked
 status::cancelled
Cancelled
 status::done
Completed
 status::in-progress
Work in progress
 status::new
New issue, not started
 status::planned
Planned for sprint
 status::review
Under review
 status::testing
In testing
 type::bug
Something is broken
 type::chore
Maintenance task
 type::documentation
Documentation
 type::enhancement
Improvement
 type::feature
New feature
 type::refactor
Code refactoring
 type::test
Testing
No Label type::chore type::documentation
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
NW olilovedani
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: UniqueSoft/APAW#25
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?